000000000001002650

Proceedings: Substation Equipment Diagnostics Conference X

1002650

Proceedings: Substation Equipment Diagnostics Conference X

1002650 Technical Update, June 2002

EPRI Project Manager B. Damsky

EPRI • 3412 Hillview Avenue, Palo Alto, California 94304 • PO Box 10412, Palo Alto, California 94303 • USA 800.313.3774 • 650.855.2121 • [email protected] • www.epri.com

DISCLAIMER OF WARRANTIES AND LIMITATION OF LIABILITIES THIS DOCUMENT WAS PREPARED BY THE ORGANIZATION(S) NAMED BELOW AS AN ACCOUNT OF WORK SPONSORED OR COSPONSORED BY THE ELECTRIC POWER RESEARCH INSTITUTE, INC. (EPRI). NEITHER EPRI, ANY MEMBER OF EPRI, ANY COSPONSOR, THE ORGANIZATION(S) BELOW, NOR ANY PERSON ACTING ON BEHALF OF ANY OF THEM: (A) MAKES ANY WARRANTY OR REPRESENTATION WHATSOEVER, EXPRESS OR IMPLIED, (I) WITH RESPECT TO THE USE OF ANY INFORMATION, APPARATUS, METHOD, PROCESS, OR SIMILAR ITEM DISCLOSED IN THIS DOCUMENT, INCLUDING MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, OR (II) THAT SUCH USE DOES NOT INFRINGE ON OR INTERFERE WITH PRIVATELY OWNED RIGHTS, INCLUDING ANY PARTY'S INTELLECTUAL PROPERTY, OR (III) THAT THIS DOCUMENT IS SUITABLE TO ANY PARTICULAR USER'S CIRCUMSTANCE; OR (B) ASSUMES RESPONSIBILITY FOR ANY DAMAGES OR OTHER LIABILITY WHATSOEVER (INCLUDING ANY CONSEQUENTIAL DAMAGES, EVEN IF EPRI OR ANY EPRI REPRESENTATIVE HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES) RESULTING FROM YOUR SELECTION OR USE OF THIS DOCUMENT OR ANY INFORMATION, APPARATUS, METHOD, PROCESS, OR SIMILAR ITEM DISCLOSED IN THIS DOCUMENT. ORGANIZATION(S) THAT PREPARED THIS DOCUMENT Thomas P. Traub, Consultant

This is an EPRI Level 2 report. A Level 2 report is intended as an informal report of continuing research, a meeting, or a topical study. It is not a final EPRI technical report.

ORDERING INFORMATION Requests for copies of this report should be directed to EPRI Orders and Conferences, 1355 Willow Way, Suite 278, Concord, CA 94520. Toll-free number: 800.313.3774, press 2, or internally x5379; voice: 925.609.9169; fax: 925.609.1310. Electric Power Research Institute and EPRI are registered service marks of the Electric Power Research Institute, Inc. EPRI. ELECTRIFY THE WORLD is a service mark of the Electric Power Research Institute, Inc. Copyright © 2002 Electric Power Research Institute, Inc. All rights reserved.

CITATIONS This document was prepared by Thomas P. Traub Substation Equipment Consultant Chicago, Illinois Principal Investigator T. Traub This document describes research sponsored by EPRI. The publication is a corporate document that should be cited in the literature in the following manner: Proceedings: Substation Equipment Diagnostics Cnference X, EPRI, Palo Alto, CA: 202. 1002650.

iii

REPORT SUMMARY Advanced monitoring and diagnostic sensors and systems are needed to provide reliable and accurate information for determining the condition of major transmission substation equipment. The tenth EPRI Substation Equipment Diagnostics Conference highlighted the work of researchers, universities, manufacturers and utilities in producing advanced monitoring and diagnostic equipment for substations. Background Utilities must continue to provide equal or better electric power service to their customers even as they continue to reduce their maintenance and operations staffs. To ensure reliable operation, the performance of major equipment in substations—including transformers, circuit breakers, instrument transformers, disconnect switches, batteries, insulators and control equipment—must be accurately and reliably monitored and potential problems detected. The tenth EPRI conference on substation equipment diagnostics provided new and followup information to the previous conferences in this rapidly developing field. Objectives • To present a forum for discussing the requirements and status of advanced diagnostic equipment and systems for substations. • To review the ongoing work of EPRI researchers, manufacturers, universities, and independent research organizations. Approach EPRI’s tenth Substation Equipment Diagnostics Conference convened on February 17-20, 2002 in San Antonio, Texas and was attended by 147 industry professionals. The four panels and 23 technical papers embraced the entire range of advanced substation diagnostics. An exhibit area displayed equipment and systems in the development, design and production phases. Key Points The tenth EPRI Substation Equipment Diagnostics Conference covered the following areas: •

For transformers: Computer diagnostics to load power transformers, new FRA developments, LTC contact coking research, on-line DGA and rate of gas generation research, utility monitoring experience, economic assessment of on-line monitoring, instrument transformer PD detection evaluation, a new approach to transformer condition assessment, forensic analysis of LTC material failures, high speed acoustic emission to detect transformer

•

•

deterioration, and transient performance of windings as a function of impedance versus frequency characteristic. For circuit breakers and other substation equipment: Using MMW for condition assessment, OCB diagnostics, life management of batteries, implementing substation predictive maintenance, wireless telemetry for assessment of equipment condition, circuit breaker operator analyzer, solutions for SF6 leak detection, diagnosis, sealing and management, For communications/data management/system integration: Tutorial on communication in substations, data standardization, integrated monitoring and diagnostics experiences.

In addition to the 23 technical papers, the Opening Session of the conference focused on how utility work will performed five years from now, the need to innovate while financial and experienced human resources are scarce, and IEEE substation monitoring standards. EPRI Perspective The conference provided an outstanding forum for interchange of information between utility engineers and those involved in the research, design, and manufacture of monitoring and diagnostic equipment. As a consequence of the interest shown and the attendance, EPRI plans to continue these conferences with the next one scheduled for February 23-26, 2003 in New Orleans, Louisiana. TR-1005915 Interest Categories Power system operations and control Substation O&M Integrated communications systems Distribution O&M Keywords Diagnostic techniques Monitoring Substations Transformers Circuit breakers Relays On-line diagnostics

ABSTRACT

Substation Equipment Diagnostics Conference X convened in San Antonio, Texas from February 17-20, 2002 to assemble, assess and communicate information on the latest diagnostic techniques, test devices, and systems for substation equipment. It focused on the latest monitoring equipment and diagnostic techniques being developed by EPRI and others in research programs, as well as the equipment and programs now available and in service by electric utilities. The conference brought together the views of researchers, manufacturers and users. The papers presented were organized under three categories: Transformers, Circuit Breakers and Other Substation Equipment, and Communications/Data Management/System Integration. Exhibit booths provided detailed information about vendor products or services for the attendees.

CONTENTS

1 PLENARY SESSION........................................................................................................1-1 Chair: Donny Helm, TXU PANELISTS: Five Years Out—Who Will Perform the Work? Rick Bush, Transmission and Distribution World Magazine The Need for Innovation Driven by Scarce Financial and Experienced Human Resources Fidel Marquez, City Public Service of San Antonio Emerging IEEE PES Standards for Substation Monitoring James H. Harlow, Harlow Engineering Associates

2 TRANSFORMER DIAGNOSTICS ....................................................................................2-1 CO-CHAIRS: Don Rose, TXU and Terry Jackson, Allegheny Power System Computer Diagnostics Used to Load Power Transformers T. J. Jackson, Allegheny Power System J. A. Uhrin, PJM L.L.C. A New Development in Power Transformer Frequency Response Analysis to Determine Winding Deformation WITHOUT the Need for Comparison to Historical Data Larry Coffeen, Jeff Hildreth, NEETRAC Mechanism of Coking in Load Tap Changers—A Status Report Russ Crutcher, Microlab Northwest Dave Hanson, TJH2B Analytical Services Leo Savio, ADAPT Corp. On-Line DGA During Controlled Energization of a 1000 MVA Autotransformer Exhibiting Partial Discharge Thomas Waters, Dan Morgan, Serveron, Inc. Profitability Assessment of Transformer On-Line Monitoring and Periodic Monitoring Jacques Aubin, André Bourgault, General Electric Canada, Inc. Claude Rajotte, Pierre Gervais, Hydro-Québec

Monitoring of Power Transformers—PSE&G Experience Larry Johnson, PSE&G USA Oleg Roizman, IntellPower, Australia Valery Davydov, Monash University, Australia Towards Maintenance-Free Load Tap Changer (LTC) II Hans-Ulf Schellhase, Ed A. Hall, Robert G. Pollock, Powertech Labs Inc. Evaluation of PD Acoustic Detection for Instrument Transformers Insulation Assessment Marceli Kazmierski, Institute of Power Engineering, Poland Ryszard Sobocki, Polish Power Grid Co. Poland Transformer Condition Assessment—A New Approach Alan Wilson, Predrag Vujovic, Doble Engineering Company John Anderson, Xcel Energy Forensic Failure Analysis: How Materials Fail in Load Tap Changers E. R. Crutcher III, Associate of ADAPT and TJH2b Detect and Locate Sources of Power Transformer Deterioration Using High Speed Acoustic Emission Waveform Acquisition with Location and Pattern Recognition Ron K. Miller, Physical Acoustics Corporation Arturo Núñez, Quality Services Laboratories Plus Overview of the Transient Performance of Coils & Windings as a Function of their Impedance Versus Frequency Characteristic Robert C. Degeneff, Mark Loose, Rensselaer Polytechnic Institute Gasses Generated From Bare Electrodes at Elevated Temperatures Immersed in Mineral Oil Observed over Ranges of Temperature and Pressure Ron K. Miller, Physical Acoustics Corporation Robert C. Degeneff, Ahmad Shahsiah, Rensselaer Polytechnic Institute Barry Ward, EPRI

3 CIRCUIT BREAKER AND OTHER SUBSTATION EQUIPMENT DIAGNOSTICS .........3-1 CHAIR: Bjorn Holm, International Switchgear Consulting Ltd Circuit Breaker Operator Analyzer Randy Cox, TXU Electric Solutions for SF6 Leak Detection, Diagnosis, Sealing and Management Mike Alfieri, Consolidated Edison Company of New York Nick Dominelli, Ian Wylie, Powertech Labs Inc. Luke van der Zel, EPRI MMW—Aiding the Task of Equipment Condition Assessment at PEPCO Scott J. Buckley, Potomac Electric Power Company

OCB Diagnostics Rick Youngblood, Cinergy Fredi Jakob, Karl Jakob, Simon Jones, Weidmann-ACTI Alex Salinas, Southern California Edison Life Management of Station Batteries Through Cell Management Gerald W. (Gary) McDermott, U.S. Bureau of Reclamation Wireless Telemetry for Assessment of HV Equipment Condition P. Duggan, M. Alfieri, P. DiLillo, Consolidated Edison Company of New York, Inc. R. Schwabe, New York Power Authority J. M. Braun, N. Fujimoto, S. Rizzetto, H.G. Sedding, Kinectrics, Inc. Implementing Substation Predictive Maintenance Jon L. Giesecke, EPRIsolutions

4 COMMUNICATIONS/DATA MANAGEMENT/SYSTEM INTEGRATION CHAIR: Michael Ingram, Tennessee Valley Authority ....................................................4-1 Data Standardization Tony McGrail, The National Grid Company, plc Integrated Monitoring and Diagnostics at Oklahoma Gas and Electric John Maio, Mark Danna, Data Systems and Solutions Lonnie Carlon, Jack Hammers, Oklahoma Gas and Electric 5 CLOSING SESSION.................................................................................................................... 5-1 6 APPENDIX: LIST OF ATTENDEES ........................................................................................... 6-1

1 PLENARY SESSION Chair: Donny Helm, TXU

Presentation to: EPRI Substation Equipment Diagnostics Conference

Rick Bush Editor-in-Chief

“Five Years Out” Who will perform the work?

Deregulation means no more business as usual Raymond James analyst Greg Haas predicts that electric utility outsourcing could grow from 30% to the 90% we now see in the telecom sector.

We’ve lived through the cutbacks

70

5 5 0 ,0 0 0

65

5 0 0 ,0 0 0

60

4 5 0 ,0 0 0

55 4 0 0 ,0 0 0 50 3 5 0 ,0 0 0

45

3 0 0 ,0 0 0

40 35

2 5 0 ,0 0 0 1992 1993 1994 1995 1996 1997 1998E1999E2000E To ta l O & M, B $

To ta l Emp lo y e e s

Employees

O&M Costs, $ in Billions

O & M a n d E m p lo y e e R o lls a t U .S . E le c tric In v e s to r O w n e d U tilitie s

Utilities fight to be to be the lowest cost producer. Non-Production Gas & Electric Cost Per Customer O&M Cost per customer*

$600 $500

Average $266

$400 $300 $200 $100 $0

*Data from 1999 FERC form 1

We are facing a degrading infrastructure

The Chicago Summer of 1999 proves: A crisis will get management investing in T&D infrastructure.

Let’s recall Chicago’s woes • July 1999, feeder trip caused power outages to 100,000 homes • August 1999 power lost to Chicago’s Loop • Another outage left Downtown Chicago in the dark

ComEd responds • 700 individuals dispatched to

inspect manholes and substations • 250,000 additional man hours expended in first six weeks

Reliability improved dramatically as ComEd coffers were drained ComEd spent in excess of 1.6 Billion dollars in massive rebuild

Vendors and Contractors more than happy to oblige •Asplundh – All tree trimming •ABB/Kenny – Fast track construction of new substations •GE/Harris – Turnkey substation monitoring and relay retrofits

Construction boomed!

SF6 Kingsbury 138kV Substation

Number of interruptions reduced

Outage duration declined

Tree trimming paid dividends

What next at ComEd? :

ComEd Integrated Contracting Strategy

: Improving the Cost Effectiveness of a Major Capital Investment Program Jim Williams Vice President Projects and Contract Management

Suggestions for Improvement „ „ „ „

Get rid of Most of Your Contracts Get rid of Fixed-Price Contracts Get rid of Most of Your Engineers Spend More Money Sooner

What to Outsource? “ALWAYS” ƒ Non-core Competencies ƒ Large Construction Projects ƒ Rarely Used Expertise ƒ Highly Technical Expertise ƒ Detail/Production Engineering

WHAT TO OUTSOURCE? “NEVER” ƒSystem Operations/Restoration ƒRoutine Maintenance ƒVery Small Projects ƒSystem Planning ƒSystem Protection

CONTRACTORS OF CHOICE MAJOR DOWTOWN TURNKEY PROJECTS ABB, Kenny Construction MAJOR SUBURBAN SUBSTATION PROJECTS GE DISTRIBUTION SYSTEM CONSTRUCTION Meade Electric, Henkels and McCoy, Aldridge Electric

CONTRACTOR OF CHOICE (continued) OH TURNKEY PROJECTS MJ Electric UD TRANSMISSION TURNKEY PROJECTS Pirelli Cable Systems NEW RESIDENTIAL CONSTRUCTION Henkels and McCoy

Others leveraged Chicago’s Woes

The American Transmission Company (ATC) announced a $37 million upgrade of Milwaukee The day after one of Chicago’s major outages!

CEO Delgado on Milwaukee: We would have been unable to take lines out of service in the summer. We would be in real trouble if certain critical circuits failed.

Milwaukee rebuilt 138kV network

SF6 Breakers installed downtown

ATC is presently outsourcing construction and O&M work.

Wade Freeman with TXU Energy Solutions is promoting the Asset Owner / Asset Manager model Based on experiences at TXU along with experiences with Eastern Electricity (UK) and United Energy (Australia) Asset Manager would be another organization in the industry – just like commercial building owners hire asset managers

Mega-contractors are forming that are willing to take on roles formerly performed by utilities •Quanta Services •Exelon Infrastructure Svcs •Duke Engineering & Svcs •Mastec

Utilities are increasingly looking to partnering, turnkey projects and outsourcing to keep costs down and reliability up •Puget Sound •ComEd •LIPA •Georgia Transmission

View from Wall street on Utility Outsourcing Energy Analyst Greg Haas

Capital Investment in Outsourcing •Deregulation Creates Opportunity & Challenges •Three Main Drivers --- Cost Reduction, Capital Expenditures, New Technology •T&D Capital Investment Still Expected To Come, But When?

What Expenses Can Utilities Outsource? •

General & Admin Comes In 2nd at $13 Billion

•

T&D Expenses Come In Third At Nearly $10 Billion

•

Customer Service & Accounts Together Make Up $6 Billion

O&M Expenditures At U.S IOUs, $ Millions $13,251 $504 $1,971 $4,000

$9,556

$57,705

Production T&D Customer Accounts Customer Service & Information Sales Admin & General

Source: EEI 1998 Statistics

Outsourcing Customer Information Systems

Survey of U.S. Investor Owned Utility CIS By Origin Number of Respondants

45

Outsourced

InHouse

40 35 30 Outsourcing Up With Deregulation In CIS Marketplace, Inhouse Apps Down.

25 20 15 10 5 0

1995 2000

2000 1995 Source: EnergyIT

Rise in Utility Capital Investment •Transmission Capital Investment Still In Infancy. •Wire Additions Not Keeping Pace With Increased Power Flow. •Existing Infrastructure More Heavily Utilized Than Ever. •T&D Spending in 2000 roughly $18 billion. •Regulatory Uncertainty Slowed Capital Expenditures – Catch Up Time Is Now. •A New Type Of Transmission -- Merchant Transmission Providers, RTOs?

Level 2 Transmission Loading Relief Events Reliability Assessment) Note: Curtailments of transactions do not occur until Level 3 (nonfirm) and Level 5 (firm). (NERC

Capital T&D Expenditures U.S. Investor Owned Utility Capital Expenditures Transmission & Distribution Segments

Catch Up Time

18000 16000 Major Blackouts

$, Millions

14000 12000 10000 8000 6000 4000 2000 0

Deregulatory Slow dow n

ƒ

The Clearest Markets

ƒ T&D Construction In Support Of New Generation ƒ Incremental T&D Outsourcing by Incumbent Utilities.

ƒ

Less Clear

ƒ Reliability Centered Maintenance ƒ Utilization Improvements, Efficiency Improvements

Outsourcing Risks ƒ Personnel Shortages ƒ Price Wars, Margin Erosion ƒ Re-regulation, Federalization, Halted Deregulation ƒ Cyclical Capital Investment Crash ƒ Distributed Generation -- Threat or Opportunity

Puget Sound Energy Is farthest along with their outsourcing strategy

Service Provider Program Scope ƒ ƒ ƒ ƒ

New Customer Construction Electric and Gas System Construction Electric and Gas O & M Construction Electric and Gas Storm Electric

Division of Responsibility Planning

PSE

Standards First Response Quality Assurance & Audit Safety

PSE PSE PSE SP / PSE

Division of Responsibility

ƒ ƒ ƒ ƒ ƒ ƒ ƒ ƒ

Engineering and Design Project Documentation Scheduling and Coordination Material Management Project Management Permitting Construction Quality Control

SP SP SP SP SP SP SP SP

„ „ „

Former Puget Employees now work for Potelco Must meet Puget standards Long-term contracts with menu of fixed unit prices

Excluded from scope ‹ Major Project Engineering ‹ Major Project Construction (Competitive Bid Work)

The Long Island Power Authority, (LIPA) is a state agency, with 62 employees. Only two work in power delivery.

LIPA manages its assets through outsourcing agreements. Agreement awarded to KeySpan Energy effective May 28, 1998 Management Services Agreement (MSA) ƒOperate LIPA’s T&D asset’s ƒTerm of 8 Years ƒManaged through performance incentives

LIPA is One of Top Three Public Power Entities by Customers and Revenues

„

2.7 million residents

Residential

953,486

Customers

Revenues $1,065,362,000

Commercial

110,024

$1,003,811,000

Other Totals

$ 1,063,510

55,767,000

$2,124,940,000

Source: 2000 audited results

MSA lessons learned Having a sole source ‘unbundled’ contract is less complex. Exclusive contracts limit competition and flexibility and don’t take advantage of “best of breed”

Preparing for the re-bid key issues ƒ ƒ ƒ ƒ ƒ

Risk and reward sharing incentives Determine core elements LIPA needs to own Determine which functions are best bundled Determine key metrics for successful MSA Level of outsourcing LIPA can successfully manage

Georgia Transmission shifts gears for growth

GTC Capital Expenditures

160

$ Million

120

80

40

0 1999

2000

2001

2002

A c tu al B u lk P r o je c ted B ulk

2003

A c tu al L o ad S e r ving P r o je c ted L oa d Se r vin g

GTC Transmission Line Projects Internal Vs. External Engineering 40 35 30 25 20 15 10 5 0 1995

1996

1997 T/L Inside

1998 T/L Outside

1999

2000

GTC Substation Projects Internal Vs. External Engineering 160 140 120 100 80 60 40 20 0 1995

1996

1997 S/S Inside

1998

1999

2000

S/S Outside

GTC Associate vs. Outside Labor Costs 14 12 10 Millions

8 6 4 2 0 1997

1998

Assoc Labor Cost

1999 Outside Labor Cost

2000

Georgia Power on building spree Ron Rhodes – Engineering and Construction Manager - Transmission

We’ve probably hired 20 engineers this year. The average age of our substation designers is 50, so we’re hiring in anticipation of attrition.

GPC still outsourcing to meet peak needs About 35% of engineering is contracted out right now

GPC outsourcing to meet peak needs Overall, we’re probably contracting out 65% of our line and substation construction.

Enron and California have only slowed deregulation Expect outsourcing to pick up speed in the years to come.

But the work remains and engineers and engineering managers are in demand. Personnel will move from cost centers to profit centers as individuals are hired by emerging services companies.

I’ve enjoyed sharing my observations with you! Feel free to contact me me anytime Rick Bush Editor-in-Chief T&D World magazine 913-967-1757 [email protected]

The Need for Innovation Driven by Scarce Financial and Experienced Human Resources Fidel Marquez, City Public Service of San Antonio

The overall theme of the conference is “How substation equipment and diagnostics will keep the lights on even with a smaller staff.” Indeed, in today’s environment, substation equipment and diagnostics is required to do more than just keep the lights on. Customer expectations are ever changing as power quality requirements are rising and tolerance for momentary outages is dropping in an environment where operating equipment at the extremes of their limits is commonplace. Continuing to add assets is becoming more challenging as NIMBY and BANANA attitudes and environmental awareness are spreading. Suppliers and Customers, both retail and wholesale, rely on substations as a “central distribution warehouse” to bring the electricity commodity to market…like it always has. This of course is nothing new, so why all the hype? The dubious experiment known as deregulation forced utilities to re-invent their businesses and become more competitive, more efficient. This even changed how we maintain and operate the electrical infrastructure. As the specter of competition became real, business models changed to reflect an unbundled enterprise. Business Units searched for ways to cut costs and maintenance was easy low hanging fruit. Some believed that this presented little short-term risk. This was at a time when maintenance was largely time-cycle driven. This may not have been a bad decision, if maintenance schedules had been current and the material condition of the equipment was adequately monitored and known. As we now know today, such was not always the case. It was quickly realized that cutting maintenance presented greater risk than anticipated. With reduced O&M budgets something needed to change. But this was just the beginning. Projects to replace aging infrastructure compete for capital with projects targeted for power plants to reduce the cost of energy, marketing programs aimed at customer retention; non-regulated offerings, and new information technology initiatives competed with projects to interconnect independent power produces and system expansion projects to meet the increased demand of the economic boom in the 1990’s. This left little, if any, capital dollars to replace an aging transmission and distribution infrastructure. Capital investments typically require some ROI payback. By their very nature, T&D replacement projects offer no incremental increased revenue and the cost

savings have a long payback, unless of course a catastrophic failure occurs is avoided, which is almost always a more expensive alternative. If reduced O&M and tight capital budgets weren’t enough, utilities went through every possible management fad at least once. Business Process Re-engineering, Reorganization, Re-alignment, Mergers, all tried to improve work flows, reduce handoffs, leverage synergies, gain scales of economy, and so on. These well-intended efforts were suppose to drive understanding of transaction costs and improve the process and alignment in order to get the maximum benefit from their resources. The most noticeable result, however was a reduction in staff…staff with many years of expertise and intimate knowledge of the system. Scare financial and experienced human resources called for innovation. We had to explore different ways to cost-effectively build, operate and maintain the infrastructure, and do so in an evolving and uncertain environment. Increasing market pressures called for increased capacity utilization, increased availability, reduced maintenance outages, reduced forced outages and, of course, reduced costs. “Better, Faster, Cheaper” has been heard from coming from every utility executive’s office in the last several years as managers and engineers pursue higher reliability, improved customer satisfaction, and lower cost. Because the resources were not available to systematically replace equipment, utility engineers, researcher and manufactures looked for ways to extend service life and develop programs and systems to actively manage the assets.

Not too long ago the term RCM surged through the industry. If it worked for the airlines it surely would work for the utilities…after all its only electricity. However, a successful RCM program is dependent on identifying what equipment is most important to operations, a thorough knowledge of failure modes and their root cause and knowing the conditions that proceed and exacerbate the failure. Most of this information is retained in shrinking tribal knowledge and it is clear that the art of substation maintenance is a need of a shot of science to take it to the next level. This level is characterized by an integrated program that combines continuous monitoring, non-invasive diagnostic testing, and regular physical inspections and provides the analysis and intelligence necessary to drive a comprehensive maintenance program. The comprehensive program will result in reduced reactive and preventive maintenance task and move toward a predictive maintenance program. In order to predict what maintenance will be needed when, the system needs to determine the condition of the equipment while in service. Which in turn, can lead to a process of determining its estimated remaining life. It wasn’t too long ago, I remember changing points and plugs as part of regular maintenance. Today there are no points and spark plugs may last the life of the vehicle.

Today’s cars are safer, faster, and more fuel-efficient. On-board computers receive signals from dozens of sensors and adjust spark and valve timing, air-fuel mixture, direct power to the wheels and so on. They also tell the technician when something is not working properly. With the exception of oil changes and wear items such as brakes and tires, today’s cares are relatively maintenance free. Earlier I stated that in today’s environment, substation equipment and diagnostics is required to do more than just keep the lights on with a smaller staff. Before too long, transformers and circuit breakers will be much like the automobiles of today. As I have described, the diagnostics must be further leveraged to also drive asset replacement and improved capacity utilization. Investment in condition monitoring equipment, diagnostic testing and data analysis will find its payback not in reduced maintenance cost or improved reliability, but in an manageable asset replacement strategy and in capital avoidance or delay, as the operating limits of the equipment is prudently extended. I am encouraged and enthused by the work and research in substation diagnostics and look forward to further improvements in reliability, reductions in cost and extension of equipment life.

Emerging IEEE PES Standards for Substation Monitoring James H. Harlow, P. E. Harlow Engineering Associates EPRI Substation Equipment Diagnostics Conference X San Antonio, Texas February 17-20, 2002 Introduction If ever there has been a Catch-22 in the electrical power business, it probably has to do with standards. Inherent in the development of a new standard is the dilemma that the standard should be prepared before the technology is developed so that everyone starts playing from the same music, but the writers can not define a standard until various, probably competitive, technologies have been tried and the best of the alternatives has emerged. Most standards fall to the second category, i.e., a reactive response to the perception that a standard is needed because industry players are individually pursuing incompatible approaches to the same problem. We can well relate to this as we lived through the onslaught of digital relays and controls from numerous vendors, each involving serial communications capability and one or more communications protocols from each of those vendors. A user might easily have had the need to install serial communications in a given substation using a half-dozen or more different protocols. That has been the case for about 15 years, and while there has been significant progress, there is still not a commonly accepted “best” alternative to the question. The Institute of Electrical and Electronics Engineers (IEEE) produces voluntary standards. As a Standards-developing Organization (SDO) it provides an important forum for industry competitors, users and others to communicate, albeit with rules and procedures in place to provide presumptive compliance with antitrust laws. Notably, the requirements imposed for balance of commercial interests, resolution of negative ballots and the appeals process are instituted to satisfy this objective. Almost one-half of the standards of the IEEE relate to power and energy. The Power Engineering Society (PES) of IEEE is responsible for writing and maintaining some 350 Standards, Guides and Recommended Practices, most of which pertain to the performance requirements of the apparatus. In the domain of substation equipment the standards fall predominantly to the Transformers, Switchgear, Stationary Battery, Power System Relaying, Surge Protective Devices and (of course) Substations Committees.

Pertinent Standards. There are a few standards issued or in preparation which deal directly with topics of monitoring of substation related apparatus. Five are identified and described: IEEE Guide for Application of Monitoring to Liquid-Immersed Transformers and Components No IEEE number yet issued As extracted from the Introduction: This Guide, still in draft form, describes most of the on-line monitoring and diagnostic methods that are in common practice, and provides additional information in the case of developing techniques. This guide addresses transformer operational parameters that can be monitored and the risks and benefits of the monitoring. It also provides specification considerations for monitoring hardware, software, and communication systems. Early in the Guide an expression is given to estimate the present value of the cost of failure over the expected life of the equipment. The equation considers the failure rate, cost of repairs, cost of outage and repair time in deriving the cost of the failure. From this may be subtracted the present value determined by using the expected lower costs of repair and repair times, if such avoids a failure. The difference between the two present worth values is the avoided cost that can be achieved by the implementation of on-line condition monitoring. Perhaps of more importance than the cost of a failure is the determination of the risk associated with a failure. High risk derives from a congruence of high probability of the event and serious consequences should the event occur. A judgement of moderate risk may result from one or the other of the probability or the consequential effect being low, even though the other remains high. From this, a risk may be qualified as Catastrophic, Critical, Moderate or Negligible. The crux of the Guide is the delineation of the parameters which may be monitored. It is observed that various continuous on-line measuring systems may include varying degrees of diagnostics as derived from an expert system which can identify problems in incipient states. Most notable of the candidate parameters to be monitored is the Dissolved Gas Analysis (DGA). The guide identifies nine gasses commonly analyzed to reveal problems which may be developing in the transformer. A very great deal of past research has convinced the industry of the value of periodic DGA testing. Continuous on-line monitoring of the Hydrogen component is common and the technology for monitoring other gases is progressing. Moisture in the oil, measured as relative saturation or concentration is recognized as an important parameter for measurement. Excessive moisture will compromise the dielectric integrity or result in accelerated aging of the transformer and must be known. Partial discharge is a precursor of dissolved gas, but an incipient problem may be detected more quickly using Partial Discharge Analysis than by waiting for the effect to show in the DGA. Electrical and acoustic methods are known for detection of partial discharge but neither has developed to the extent desired as an unambiguous PD measurement.

Perhaps the oldest monitoring of transformers is simply the measurement of top oil temperature. An elevated temperature is expected with overloading, but can also be attributed to failure of pumps or fans. Of even more consequence is the measurement of the hot-spot winding temperature which is commonly modeled but is the subject of more recent research using fiber optics for direct measurement. Perhaps so obvious as to be dismissed for recognition is the monitoring of load current and voltage. This knowledge in conjunction with temperatures can produce significant information about the state of the transformer. A common means of cross checking for moisture or contaminants in the insulation, which produce dielectric loss, is the measure of the insulation power factor (or dissipation factor). This test is routinely performed when a transformer is deenergized for maintenance. Proper cooling of the transformer is wholly dependent on the operation of the pumps and fans which circulate the oil and force air across the radiators. These are readily monitored for operation by measuring the current drawn by the motors and correlating that with the temperature control of the cooling system.

The Guide recognizes the particular concerns associated with monitoring of the Load Tap Changer (LTC). Some of the same thoughts associated with the transformer per se are relevant for the tap changer. The LTC, being a mechanical device, is also a candidate for other considerations such as vibration monitoring and signature analysis. The Guide identifies, without elaboration, 76 candidate parameters which, if monitored, may be useful in the diagnostics of the transformer condition. The Guide then continues with a detailed description of possible monitoring systems and equipment. This entails description of the likes of sensors, data acquisition units, data processing, hardware specifications and the selection of communications protocols for on-line monitoring equipment, identifying DNP3 and IEC 608705 as leading candidates. The Guide also recognizes UCA2.0 as an emerging possibility. A very significant portion of the Guide continues with application thoughts pertaining to power transformers, instrument transformers, bushings and load tap changers. The final section deals with Data Application treating topics of sampling criterion, statistical method, data analysis, trending, expert system diagnostics and partial discharges.

IEEE Guide for the Selection of Monitoring for Circuit Breakers IEEE Std C37.10.1-2000 This Guide provides guidance in the selection of monitoring for high-voltage circuit breakers. The introduction advises that monitoring for a particular circuit breaker is very dependent on the circuit breaker technology, age of the circuit breaker, details of the specific application, and the risks associated with the various failures possible with the circuit breaker and its many associated power or protection and control and other support components. A decision making sequence is followed in the Guide:

• Perform a Failure Mode and Effects Analysis (FMEA) to identify the possible failure modes. This will also reveal monitoring options appropriate for observing the condition of the circuit breaker in question. • Conduct a risk assessment to quantify the risk associated with each identified failure mode. This will lead to a cost-benefit analysis. • Make the decision to implement continuous or periodic monitoring based on the earlier studies. In some cases this will be evident; other cases, will require that the decision be based on a reasoned judgement. The Guide gives a very succinct description of a FMEA. This may be considered as a tutorial for those not well initiated in the procedure. Using the FMEA teachings, the bulk of the Guide presents 17 Tables which delineate failure modes, possible effects and causes of the failure. The Tables also include ideas on the failure characteristic and possible monitoring options. As just one example of the FMEA, consider where the circuit breaker opens but fails to interrupt the circuit: Failure Mode: • Circuit breaker opens but fails to interrupt. Failure Effect: • Fault or load current is not interrupted and the circuit breaker interrupter has a major failure. Possible Failure Causes - Failure Characteristics - Monitoring Options: • Oil contamination - Slowly over time - Oil dielectric • Low gas pressure or density (air or SF6) - Slowly over time - Gas pressure or density • Loss of vacuum - Instantaneous or gradual - Periodic vacuum integrity overpotential test • Insufficient contact opening - Instantaneous or gradual - Contact travel • Arc chute failure - Slow or sudden mechanical failure, heater failure, contamination, or connection failure - Visual inspection, partial discharge, and heater or temperature • Puffer failure - Slow or sudden – [Not specified]

• Mechanical failure - Slow or sudden - Mechanism position and auxiliary contacts with respect to current flow and opening signal • Misapplication or other situation beyond circuit breaker capability - Developing system generally changes with reconfigured system or new switching duty - System fault level and conditions, especially during short circuit interruption and breaker operation • Misapplication or other situation beyond circuit breaker capability - Interruption under improper system application conditions - Periodic review of system fault levels & power system disturbance recorder (including oscillographs and digital fault recorders)

Other possible failure modes are described with equally comprehensive analysis. In a manner very much the same as that reported for the monitoring of transformers, above, it is noted that a risk assessment is required where risk is derived from the two factors of 1) the probability of the event occurring and 2) its consequence. Identical to that reported for the transformer case, this Guide delineates the risk as Catastrophic, Major, Moderate or Negligible with a description of each possibility in terms of financial impact, possible bodily injury and the harm caused, with particular concern for ecological damage.

The Guide then very effectively presents a second tutorial, this dealing with the cost-benefit analysis. As the event may be expected to occur Frequently, Occasionally, Infrequently or Improbably, this section can help a user of switchgear equipment consider and justify the implementation of monitoring and diagnostics programs based on the analysis of costs and benefits. The same cost of failure basis reported in the transformer Guide is in this switchgear Guide. Continuing the tutorial nature of the Guide, two very detailed and concise examples are presented. The examples present cases arguing for and against monitoring based on an economic justification.

IEEE Guide For Selection and Use of Battery Monitoring Equipment in Stationary Applications PAR 1491/D8 - April 2000 To be very precise, the PES Stationary Battery Committee isn't doing any work in this area, but SCC29, which is closely allied to the Stationary Battery Committee, has the referenced project. The Guide identifies operational parameters that may be observed by lead-acid and nickelcadmium battery monitoring equipment used in stationary applications, and the relative value of such observations. The Guide identifies nine parameters which may be considered for manual monitoring: • Voltage of each cell and total battery voltage • Direct Current • Alternating Current Ripple • Temperature • Interconnection Resistance – in micro-ohms • Specific Gravity • Visual Inspection – look for physical defects, sedimentation, plate condition, sulfation and leakage of electrolyte. • Internal ohmic measurement • Insulation breakdown – Resistance from a terminal post to the tray or rack; accomplished at 500 VDC. Remote Monitoring Systems can perform most of the functions listed above on an automatic, unattended basis. A very significant portion of the Guide is devoted to a comprehensive 1) Description, 2) Purpose of Monitoring and 3) the Limits associated with the various monitored parameters. A treatise is provided on environmental factors and various sensor technologies which may be employed in the monitoring apparatus. Hints are offered regarding operational software, alarms and reports. The Appendix to the Guide offers some thoughts on the means of communicating the monitored data.

IEEE Trial Use Guide For The Detection Of Acoustic Emissions From Partial Discharges In Oil-Immersed Power Transformers PC57.127 - April 2000 C57.127 is a published Guide, currently in revision. The Guide may be considered to be a detailed study for acoustic monitoring of partial discharges (PD) in a transformer and therefore an elaboration on that aspect of the transformer monitoring Guide first reported in this paper. The Guide is predicated on the recognition that each partial discharge occurring within the insulation produces a low amplitude mechanical pulse which propagates to the tank wall where it can be detected by an appropriate sensor. The count rate of such pulses is indicative of the intensity of PD activity. The Guide instructs the reader regarding responses in which “continuous” readings are obtained, as contrasted to other responses producing “sporadic” readings. The mechanics of accomplishing the monitoring is treated in two Annexes, “Instrument Calibration” and “Calibration of Transducer and Preamplifier.” There is no material relating to continuous on-line monitoring; the provisions of the Guide are directed to testing in the field, but are certainly also applicable to laboratory testing.

IEEE Recommended Practice for Data Communications Between Intelligent Electronic Devices and Remote Terminal Units in a Substation IEEE 1379-2000 Surely not to be overlooked is the aspect of communicating of the monitored information. A major endeavor of this aspect was undertaken by the Substations Committee which developed IEEE 1379 to put the IEEE stamp on the protocols which most nearly satisfied the requirements as defined by the members. The RP designates two existing protocols, DNP3 or IEC 60870-5101, as preferred. With DNP3 being, de facto, the protocol of primary choice throughout North America. Another protocol, IEC 61850, is being developed and is recognized to be a viable candidate when it is complete. This protocol is derived in large part from efforts under the auspices of EPRI to develop UCA2.0 where UCA relates that this is intended as the Utility Communications Architecture. Much of the definition for UCA was produced in a document entitled Generic Object Models for Substation and Feeder Equipment (GOMSFE) which was the basis of a report at this conference in 1998. GOMSFE then became the foundation for Data Object definitions in IEC 61850. The expectation is that IEC 61850 will be approved in 2003. The matter of communications is a segment of this conference being itself the topic of concentrated discussion. It is not useful to attempt to further elaborate the topic in this report.

Summary This report has attempted to identify the principal activity in the realm of standards relating to monitoring of substation equipment. There may certainly be other standards with which I am not acquainted, but those reported can at least be said to give a flavor of the work involved. Standards writing is an ever continuing process. All IEEE standards must be revised or reaffirmed every five years. Those not so acted upon are withdrawn. All PES Technical Committees who write these standards are continually seeking new expertise to participate in the work. Wherever possible, I hope you individually will consider how you may be an active part of it. Acknowledgement Materials which were used to compile this paper are gratefully acknowledged as having been provoded by Donald Chu, Andre Lux, Bill Bergman, Jim McDowell, Jack Harley and Ron Farquharson.

2 TRANSFORMER DIAGNOSTICS Co-Chairs: Don Rose, TXU and Terry Jackson, Allegheny ower System

COMPUTER DIAGNOSTICS USED TO LOAD POWER TRANSFORMERS

T. J. JACKSON ALLEGHENY POWER J. A. UHRIN FORMALLY OF ALLEGHENY POWER PJM L.L.C.

SUBSTATION EQUIPMENT DIAGNOSTICS CONFERENCE SAN ANTONIO TEXAS FEBRUARY 17-20, 2002

COMPUTER DIAGNOSTICS USED TO LOAD POWER TRANSFORMERS T. J. JACKSON ALLEGHENY POWER J. A. UHRIN FORMALLY OF ALLEGHENY POWER PJM L.L.C.

Since 1974 Allegheny Power (AP) and the various subsidiaries that preceded AP have been using computer programs to develop the loading on power transformers. The first computer program was based on the IEEE guidelines for loading power transformers that were in effect at the time. The software output consisted of an MVA rating for the power transformer at summer (30° C ambient) and winter (10° C ambient) temperatures. The acceptable transformer loading or rating was developed for the transmission facilities when it was placed in service and that value was left as the rating for the life of the transformer. Over the next 20 years the program worked well. However in 1996 and the introduction of deregulation in the electric utilities, power sales into and across multiple transmission system became common. As a result some systems experienced heavier than normal loading on transmission system facilities due to scheduled and unscheduled power flows. Competition between generation suppliers forced many companies to limit capital budget project expenditures and required a detailed economic analysis to justify the funds. The result was limited pool of monies for system upgrades. Existing equipment use was called upon to carry maximize their usefulness. Because of this it became imperative that software be developed to assist new engineers in calculating loadability values that would reliably maximize transmission facility loading while not jeopardizing system facilities and electric service reliability. Rather than rewrite the existing 25+-year-old transformer loading program, AP decided to use the EPRI PT Load computer program for the calculation of transformer loading. So in 1997, AP began using PT Load to re-rate some of the power transformers that reside on the AP Transmission System. That computer program was PT Load, Version 4.1, which turned out to be very cumbersome and non-user friendly. In 1998, EPRI began to rewrite the PT Load program and in 1999 issued Version 5.0. This version of the software, PT Load Version 5.0, was developed on Windows based platform that was very userfriendly. Further enhancements of the software by EPRI resulted in PT Load Version 5.1. Allegheny Power used PT Load Version 5.1 on two separate occasions during 2001 to assist real-time operations. On both occasions, power transformer ratings in two different substations were reevaluated and limited loading on transformers and protect them from thermal overloading. Allegheny Power has estimated that during this past summer a savings of $1.7 million was realized over a three-week period, during the month of August. A summary of each application of the PT Load software is provided below.

Kammer T200 On May 22, 2001, a typical spring day (70° F) the load on the Kammer transformer was fairly low 600MVA. At approximately 7:00AM, loading increased to 1200MVA and subsequently increased to 1530MVA by noon. As information, the Kammer T200 Power Transformer is a 765-525kV, 1500MVA (3-1 Ø 500MVA each) FOA rated transformer. The transformer loading varied slightly over the next several hours until approximately 1430 when the transformer tripped itself out of service due to a pressure relief operation. When substation personnel arrived on site, they observed that oil had been expelled out of the pressure relief device on one of the three single-phase transformers (Transformer Unit #1). Allegheny’s relay scheme is set up such that upon an operation of the pressure relief device the transformers automatically removes itself from service. These transformers are nitrogen blanketed positive pressure units. The substation personnel on site recorded the following temperature gauge readings: Unit 1 Common winding Series winding Top oil

124° C 110° C 80° C

Unit 2 125° C 123° C 98° C

Unit 3 130° C 110° C 90° C

A review of the transformer loading data, which was gathered by Substation Maintenance personnel and analyzed by Substation Engineering led to some confusion over the cause for the transformer tripping. Substation personnel were puzzled because Unit 1, the transformer that expelled oil, had lower temperature readings than the other two transformer units on site. Substation personnel came to the conclusion that there was too much oil in the transformer units. Personnel on site arranged to drain the excess and the transformers were returned to service on May 24, 2001. On May 30th Engineering was asked to review the incident. Realizing that the ambient temperature was only 70 °F and that the load was well below its calculated (old method) rating of 1590MVA, an investigation and study was initiated to determine what caused all the gauges to read so high and also determine why the oil was expelled. Using the PT Load software, actual system data that included transformer loading and ambient temperatures for the previous 48 hours, were inputted into the computer program. Results from the software indicated that the common winding temperature should have read 103° C and the top oil 50° C (see Figure 1A). A difference of 20° C and 30° C respectively from the actual system readings recorded by the substation personnel. An analysis of the data and the results from the PT Load software determined that since both temperature values were so far off, the problem had to be with the transformer and not the transformer gauges. To diagnose the problem, the PT Load program was re-run using the actual loads and ambient temperature but then varied the nameplate rating, the load losses and oil thermal time constant of the transformers, until the output top oil read 80° C. The program re-runs continued with actual load and ambient temperature using the revised nameplate rating and load losses until the winding temperature equaled 124° C. In all cases the new nameplate rating was approximately

427 – 435 MVA per unit. Refer to figure 1B. Remembering that the effective MVA rating of a power transformer is determined by the square root of the operating cooling divided by the total cooling and allowed PT Load to determine the rating of the transformer. The PT Load software calculated a transformer rating of 427 MVA rating rather than the nameplate rating of 510 MVA per unit. Engineering determined that approximately in order for this to happen physically, 1/3rd of the transformer cooling was not functioning properly. AP Operations was informed of the transformer loading concern and the transformers’ inability to cool itself. Operations lowered the transformer rating immediately. AP Substation Maintenance personnel made a visit to the substation and upon inspection of the transformers it was obvious that the cooling on this FOA transformer had not been properly cleaned. Throughout this analysis, the calibration of the gauges was also questioned. Before AP jumped to the conclusion as a result of using PT Load, an infrared test was also done on the transformer units. Results from the infrared test found that on average the difference between input and output of the nine coolers on the transformers was only 1° to 2° F. The coolers were partially disassembled and pressure washed. The gauges were recalibrated. Another infrared test was redone and the difference between input and output of the cooling was raised to roughly 10°F on each cooler. By the end of June the transformers maintenance was complete and allowed Operations to re-elevate its nameplate rating back up to 1530 MVA. This exercise showed to AP Operation and Substation Maintenance personnel the power of PT Load not only a transformer loading program but that it could also be used as a diagnostic maintenance tool. Doubs Power Transformers On July 25th two of the four Doubs 500-230kV, 210/280/350 (392) MVA power transformers went into winding temperature alarm (120°C). This prompted the Transmission System Operator to open a 230kV line in order to relieve loading on the power transformers and thus removing them from alarm. The transformers had just reached their maximum rated load capability when the alarms were received by the Operations Center. Substation personnel were dispatched to the substation and temperature gauge readings were recorded. Subsequent temperature readings were taken on all the gauges on each transformer over the next four hours. See figure 2 for a listing of those readings. AS a result of these reading, Allegheny Power Substation Engineering developed the following set of observations: 1. All power transformers at Doubs are the same continuos MVA rating. Meaning that the gauges should be within +/- 5° C of each other. 2. Three of the four transformers were above the winding temperature alarm of 105° C, but only two went into alarm. 3. A fourth transformer had a 10° C difference between winding and liquid temperature.

As a result of the observations and the differences between the actual transformer gauges readings and the alarms being received, Operations asked Engineering to investigate why there were differences and determine the correct loadability limits necessary to prevent over load on the four transformers. Figure 3 compares the PT Load computer calculated readings of all the gauges versus what was a Substation personnel actually read on July 25th. Again, Substation Engineering developed a set of observation, found below as a result the PT Load software tests : 1. The liquid temperature readings comparisons were reasonable. 2. The winding temperature gauge actual readings were high on units T1, T2, T3 3. The winding temperature gauge actual reading on T4 appeared to low. 4. As the transformer loading and ambient temperature subsided later in the day, the winding temperature and PT Load output seemed to correlate. 5. Of all the gauges T1’s seemed to be the worst when it came to calibration. The conclusion of the analysis was that the winding temperature gauges needed recalibrated on all four-power transformers at Doubs. Two weeks later on August 7, 2001, during a hot summer day with heavy loading expected on the power transformers, units T1 and T2 went into alarm again. Operations requested the assistance of Substation Engineering and use of the PT Load software to aide in the calculation of acceptable transformer ratings. After some additional analysis, Engineering and Operations decided to use the winding and liquid temperature readings from the PT Load software as the alarming points to load the transformers. Since these transformers play such a critical role on the Allegheny Power System, it was decided to recalibrate the gauges on T1 as soon as possible. The Substation Maintenance personnel indicated the work could be done in a six-hour period from midnight until 6:00 AM. The result of the maintenance work was that: 1. The winding and liquid temperature gauges were re-calibrated. Re-calibration confirmed that the gauges were reading 4° C to high especially on T1. 2. The set point (temperature) for the cooling stages to come on was readjusted. 3. The winding temperature alarm was corrected and set at 105° C. To confirm the value placed upon the PT Load software, additional, what we called follow up software testing was done. Figure 4 shows the comparison of the PT Load computer output for winding and liquid temperatures before and after the recalibration. The table of results and a comparison of the data show a much better relationship between the real-time data and the results of PT Load. Even with the few times the AP criteria of +/- 5° C was exceeded. As a

result of the and their experience and trust developed in PT Load, AP intends to install electronic gauges on the four power transformers at Doubs so the readings can be remotely monitored and the data can come directly into their EMS by way of SCADA. By having this software automated to the point that real-time values will be used as input data, the PT Load program will be used and executed by trained Operations personnel in order to monitor transformer temperature alarms and compare program results and actual values. Operations and Substation Maintenance have through these two real life experiences, gained faith and trust in the program and believe that over time the program can and will become a useful tool for both the Operation Center and Substation Maintenance.

Kammer T200 765-500kV

3-1Ø Units 500MVA each Installed 1973

Actual Readings UNIT 1 2 3

May 25, 2001

HOT SPOT 124 125 130

HOT OIL 80 98 90

PT Load Calculation UNIT 1

HOT SPOT 103

HOT OIL 50

Figure 1A

Kammer T200 PT Load Reading with Adjustments UNIT 1

HOT SPOT 124

Unit Rating 427MVA Adjusted No Load to Load Values Top Oil Over Ambient Conclusion: 1. Gauges Re-adjusted 2. Coolers Cleaned Figure 1B

HOT OIL 57

Doubs Power Transformers Actual Readings 7-25-01 TIME 16:25 H.S. SERIES H.S. COMMON H.O.

T1 121 96 77

T2 110

17:45 H.S. SERIES H.S. COMMON H.O.

120 96 78

109

18:45 H.S. SERIES H.S. COMMON H.O.

119 94 78

109

20:10 H.S. SERIES H.S.COMMON H.O.

112 90 75

103

70

70

70

68 Figure 2

T3 112 108 70

T4 86 90 76

106 110 70

85 91 74

105 109 70

86 85 72

100 103 67

82 80 70

Doubs Power Transformers 7-25-01 TIME

T1 Actual-Ptload

T2 Actual-Ptload

T3 Actual-Ptload

T4 Actual-Ptload

16:25 H.S. H.O.

121 114 77 75

110 70

104 72

112 70

107 70

90 76

107 69

17:45 H.S. H.O.

120 116 78 77

109 70

106 73

110 70

109 71

91 74

108 70

18:45 H.S. H.O.

119 116 78 77

109 70

105 73

109 70

108 71

86 72

108 70

20:10 H.S. H.O.

112 111 75 75

103 68

100 71

103 67

102 69

82 70

102 68

Figure 3

Doubs’s T1 Temperature Readings

Time

8-7-01 Actual-Ptload

8-8-01 Actual-Ptload

8-9-01 Actual-Ptload

12:00 H.O. H.S.

67 100

64 93

62 94

63 91

66 96

67 98

13:00 H.O. H.S.

69 103

66 96

65 102

66 96

69 106

70 102

14:00 H.O. H.S.

71 107

68 98

68 106

70 103

72 110

72 106

15:00 H.O. H.S.

73 109

70 101

70 109

72 106

74 117

75 111

16:00 H.O. H.S.

75 115

72 103

73 111

74 107

75 118

76 111

Figure 4

A New Development in Power Transformer Frequency Response Analysis to Determine Winding Deformation WITHOUT the Need for Comparison to Historical Data [The Objective Winding Asymmetry Test] Larry Coffeen Jeff Hildreth NEETRAC High Voltage Laboratory A Center of the Georgia Institute of Technology

Abstract A new development in power transformer Frequency Response Analysis (FRA) was presented at the EPRI Substation Equipment Diagnostics Conference IX by these authors [3]. During the past year we have developed an additional enhancement to the off-line test technique called the Objective Winding Asymmetry (OWA) test . The OWA test is based on the same spectral density building blocks but used in a different manner to reveal new information. Our FRA test data indicate that there is a certain amount of symmetry between the windings on a given 3 phase transformer in good condition. There is also a certain amount of symmetry between “identical sister” transformer windings. The same objective algorithm for comparing a present and past winding frequency response can be used to indicate the amount of asymmetry between phase windings and therefore indicate the winding which has significant deformation. Four case studies are presented to show the viability of the new technique along with an objective definition of winding asymmetry. All cases are resolved without the use of previous FRA data. Background Summary Many utilities are already using some form of off-line FRA on new and existing transformers because it is well-known to be sensitive to winding distortion. No other known technique is as sensitive to winding deformation (coils, layers, turns, leads etc.). The major difficulties associated with all of the existing traditional FRA techniques are associated with: (a) the interpretation of the data presented and (b) what the acceptance criteria should be. NEETRAC has made significant progress in resolving these issues with the newly developed Off-Line Test Set [3].

The NEETRAC Off-Line FRA Test Set is based upon patent pending software which uses Spectral Density Estimate (SDE) building blocks to form the transfer function H(f) [4]. The advantages of the SDE transfer function with the associated coherence and random error calculations over the traditional transfer function are as follows:

• • • • • •

Includes a built-in estimation of confidence in test accuracy and repeatability Detects usable bandwidth Built-in noise rejection Only 10 test pulses are required per test configuration SDE’s are enhanced when input pulses are slightly different or very different in shape The hardware requirement for the input pulse box is greatly simplified since precise input pulse shape repetition and precise timing between input pulses are no longer required.

NOTE: The SDE equations and the steps of the NEETRAC FRA technique development are outlined in Reference [3]. NEETRAC Software for Interpretation of FRA Test Data / Acceptance Criteria A Weighted Normalized Difference (WND) number is calculated with a software algorithm to indicate the condition of change between a previous and the present FRA test. The WND number will indicate a “green”, “yellow”, or “red” zone of transformer winding condition. The NEETRAC Off-Line FRA Test Set is designed so that a single condition number or WND number is available for every comparison of two transfer functions. Typically, this would be the previous FRA test compared to the present FRA test for a given test configuration on a transformer. The key points of the WND calculation are as follows: • • • •

The difference of H1(f) & H2(f) is computed at each frequency Each data point is normalized Each data point is weighted according to the error function at that frequency WND = a constant times the average of the weighted values

In addition, the WND numbers are divided into three ranges as follows: • • •

The “Green Range” is represented by a WND number of 5 – 25 which indicates “No Detectable Change” to a “Minor Change.” The “Yellow Range” is represented by a WND number of 26 – 75 which indicates “Some Change.” The “Red Range” is represented by a WND number of 76 – 300+ which indicates a “Significant Change” to a “Major Change.”

These ranges are based on a test database of forty-two transformers (12 distribution transformers and 30 substation power banks) tested over a two year period. Nine of these transformers (25kVA Pad Mounts) were FRA tested before and after short circuit testing and then torn down to match the extent of winding damage to the WND number result. There were also many repeat

tests on the same transformers to determine the effects of transportation, heat, oil processing, and test repeatability (precision) [3]. The Objective Winding Asymmetry Test

Introduction NEETRAC spent about a year doing routine FRA testing on substation power transformers to gather a database for the traditional present-vs-past test comparisons. The Transmission Service Engineers of the Southern Company (SoCo) begin to ask NEETRAC if we could test a transformer without any previous test history and make a judgment on winding deformation. So we began to develop the same objective FRA method used for present-vs-past transfer functions (TF’s) to compare differences between “H” windings and differences between “X” windings on a 3 phase transformer. We discovered that the windings of the same voltage are not identical, but, in most cases, they are similar to the extent that any significant winding deformation will produce a computed Objective Winding Asymmetry (OWA) several times larger than normal. There have been 25 power banks tested for OWA in the past year. Three of these transformers had significantly large OWA on a specific “H” or “X” winding. One was verified by internal inspection. This same transformer had passed all of the normal tests including Power Factor, Excitation, TTR (transformer turns ratio), and DGA (dissolved gas analysis). Another OWA finding was verified by a TTR test which indicated a shorted winding, and the third large OWA was eventually verified by a transformer replacement in a power plant adjustable speed drive application. The suspect transformer with a high “H” winding OWA passed all the normal tests plus additional tests which included Two Times Induced Voltage, Lightning Impulse, and a Full Load test. In two out of the three cases, the NEETRAC OWA test was the only indicator of the winding deformation. Definition of OWA using WND Numbers and OWA Test Results The WND numbers are calculated in the same manner as presented earlier in this document and in Reference [3]. The difference is that each “H” winding is compared to the other two “H” windings, and similarly, each “X” winding is compared to the other two “X” windings by OWA. The OWA is given in percent and is defined by the average of the two high WND numbers divided by the lowest WND number for the 3 separate winding comparisons. Then subtract one from the result and convert to percent. See Figure 1 for the OWA test results on a recently reworked 230/115 kV auto bank, 280 MVA, manufactured in 1987. Note that there is no significant asymmetry detected on this transformer. The OWA can also be calculated for winding coupling comparisons between insulated windings on transformers with more than one insulated winding per phase (I.E. non-auto banks such as delta-wye, wye-wye, or delta-delta transformers). The OWA for these coupling comparisons can be used as additional information to support the OWA for the direct winding comparisons. See Figure 2 for the OWA test results on a 46kV/480 volt, delta-wye, 4.7 MVA, manufactured in 1999. Observe that the OWA coupling comparisons also point to the same leg of the transformer on which the asymmetrical X1 is located.

Objective Winding Asymmetry 11-7-01 Test 230/115kV auto, 280MVA Direct WND# Comparisons H1_H0X0 vs H2 – H0X0 112.0 H2_H0X0 vs H3 – H0X0 138.2 H3_H0X0 vs H1 – H0X0 166.9 H3_H0X0 = 36.2% Asym. COMMENT: H0X0 is nearest H3

Direct WND# Comparisons X1_H0X0 vs X2_H0X0 178.8 X2_H0X0 vs X3_H0X0 178.1 X3_H0X0 vs X1_H0X0 156.1 X2_H0X0 = 14.3 % Asym.

NO ASYMMETRY FOUND Application Guidelines / Acceptance Criteria

• •

“H” windings without deformation are typically less than 100% asymmetrical “X” windings without deformation are typically less than 50% asymmetrical OWA Test Results Figure 1

Objective Winding Asymmetry 9-6-01 Test 46kV/480v, ∆-Y, 4.7MVA Direct Comparisons H1H2 – H2H3 H2H3 – H3H1 H3H1 – H1H2

WND# 129.4 200.8 143.1

Coupling Comparisons H1X0 – H2X0 H2X0 – H3X0 H3X0 – H1X0

WND# 132.1 76.5 130.9

Direct Comparisons X1X0 – X2X0 X2X0 – X3X0 X3X0 – X1X0

H3H1 = 32.9% Asym. H1X0 = 72.9% Asym.

WND# 53.2 46.2 85.9

X1X0 = 50.5% Asym.

COMMENT: TTR indicated short in X1X0 winding. H3H1 has the long internal lead.

No obvious winding displacement on internal inspection. OWA INDICATES X1X0 DEFORMATION ON COUPLING & DIRECT COMPARISONS.

Asymmetry Guidelines / Acceptance Criteria • • •

A long internal lead on one “H” winding (see nameplate) can cause some asymmetry for that winding. Coupling Comparisons for small MVA transformers should be much less than 50% Asymmetrical. Small MVA or low voltage “X” windings should be much less than 50% Asymmetrical.

OWA Test Results Figure 2

OWA comparisons can also be made between identical sister transformers, single phase generator step-ups, or single phase reactors etc. See Figure 3 for the OWA test results on 3 single phase 500 kV reactors. NEETRAC had been advised that phase 2 of the three reactors on a 500 kV transmission line from GA Power’s Plant Hatch had relayed out on neutral unbalance current. The test engineers had checked the relays, and asked NEETRAC to do an FRA test on the unit to help in the investigation. Power Factor and DGA tests were also completed. The reactors were “identical sisters,” manufactured in 1982, so an OWA test was performed using all three units. Note the results in Figure 3 show no indication of winding asymmetry. Upon further investigation, the problem was determined to be in the relaying and control circuits. The phase 2 reactor has successfully been back in service for over 8 months since the occurrence.

Objective Winding Asymmetry 6-5-01 Test 3-500kV, 50MVA Reactors Direct Comparisons Phase 1 vs Phase 2 Phase 2 vs Phase 3

WND#

Phase 3 vs Phase 1

90.8

Direct Comparisons

WND#

61.7 49.6

Phase 1 = 53.7% Asym. COMMENT: NO INDICATION OF A PHASE 2 (OR 3 OR 1) WINDING DISPLACEMENT

Application Guidelines / Acceptance Criteria •

“H” windings without deformation are typically less than 100% asymmetrical.

OWA Test Results Figure 3 One of the most displaced windings discovered to date was in the early stages of the OWA development. As a result, the OWA direct comparison test of the “H” windings was omitted on this particular transformer. However, the data actually obtained was sufficient to point to the displaced winding. This transformer was in Georgia Power’s spare storage waiting to be placed in service. This power bank had passed all of the normal tests including Power Factor, Excitation, TTR, & DGA, and was being prepared for transportation to service. An internal inspection was performed when the oil was removed and immediately a considerable amount of winding distortion was discovered. The manhole covers were replaced and NEETRAC was asked to perform the OWA test on the unit without oil. The transformer was a 115/12 kV, 4.2MVA, delta-wye bank, manufactured in 1966. The results of the OWA test are shown in Figure 4. The pictures taken looking down through the top manhole to indicate some of the X2X0 winding distortion are shown in Figure 5 and Figure 6. Closer inspection was not possible since the unit went directly to be scrapped without being untanked.

Objective Winding Asymmetry 3-6-01 Test 115/12 kV, ∆-Y, 4.2 MVA Coupling Comparisons H1X0 – H2X0 H2X0 – H3X0

H3X0 – H1X0

WND# 58.7 77.0 81.0

Direct Comparisons X1X0 – X2X0 X2X0 – X3X0

WND#

X3X0 – X1X0

250.2 230.9 54.2

H3X0 = 34.6% Asym. X2X0 = 344% Asym. COMMENT: Power Factor Test, Excitation Test, TTR Test, DGA were normal. Internal inspection revealed displaced “X2” winding. OWA INDICATES LARGE ASYMMETRY FOR X2X0 WINDING.

Application Guidelines / Acceptance Criteria • •

Coupling Comparisons for small MVA transformers should be much less than 50% Asymmetrical. Small MVA or low voltage “X” windings should be much less than 50% Asymmetrical.

OWA Test Results Figure 4

Loose Packing / Insulating Materials

Top Inside View of Center Leg Coil Assembly Loose Materials Figure 5

Gap indicates loose coil

Loose Blocking Figure 6

Comments on the Practical Application of OWA Refer to Figure 1, for example, and notice that the differences between the windings is significant, but they are equally different from each other. So the Asymmetry is still low. Asymmetry is defined in the OWA test to point out the outlier, not the ones that are different or not identical. It is true that the windings which are close to being identical will also show little asymmetry, but the windings can be very different as long as the amount of difference is about the same. This definition of asymmetrical has worked well on all 25 of the substation power transformers tested to date with the OWA test.

Objective Winding Asymmetry Test Advantages over the Traditional Present-vs-Past FRA Test Method • • •

•

OWA can indicate winding deformation immediately with NO PAST FRA TEST HISTORY REQUIRED. Immediate results are very valuable to today’s transmission service engineers. The effects of oil & paper aging, test temperature differences, & moisture content at time of test are canceled out because these parameters are constant during a test. Many times a transformer must be transported to a destination with different test temperatures due to climate or time from de-energization etc. and filled with newly processed oil versus the aged oil at the time of initial testing. For this case the OWA test comparisons between the initial and final bank destinations will be a better and more consistent indicator of winding movement during the move. Initial OWA tests indicate that the test works equally well without oil in the transformer which is a considerable operational advantage since a transformer usually cannot be transported filled with oil.

Conclusions from Initial OWA Test Development on 25 Substation Transformers •

The NEETRAC Objective Winding Asymmetry (OWA) test technique can be used effectively to determine substation power transformer winding deformations WITHOUT past FRA test history.

•

The interpretation of the data presented and the acceptance criteria for the data are derived objectively in software without test personnel intervention.

Future Work NEETRAC will be working with member and participating companies to develop the On-Line FRA technique, based in part on the unique properties of Spectral Density Estimates. Patent protection has been filed by Georgia Tech Research Corporation for both the NEETRAC OffLine FRA test set [4] and for the On-Line FRA Technology [5] developed. Acknowledgements The authors would like to recognize the visionary commitment of two of NEETRAC’s members, The Southern Company and South Carolina Gas and Electric, who made the development of the Off-Line FRA test set a reality. The authors would particularly thank Mr. Don Cantrelle and Mr. Danny Bates of SoCo for their support of the OWA development. References 1. Julius S. Bendat, and Allan G. Piersol, “Engineering Applications of Correlation and Spectral Analysis", Second Edition. John Wiley & Sons, Inc. 1993. 2. James E. McBride, and Larry T. Coffeen, “The Application of Spectral Density Based Estimates in Processing Digital Records from High Voltage Measurements”, Proceedings of International Symposium on Digital Techniques in High-Voltage Measurements, Toronto, Canada, October 28-30, 1991. Sponsored by CIGRE Study Committee 33, WG 03, IEEE PES PSIM Committee. 3. Coffeen, L., Hildreth, J, “A New Development in Power Transformer Off-Line & On- Line Frequency Response Analysis,” Proceeings of the 2001 EPRI Substation Equipment Diagnostics Conference IX, Feb. 19-21, 2001. In New Orleans, LA . Paper presented by L. Coffeen. 4. U.S. Utility Patent Application, “System and Method for Off-Line Impulse Frequency Response Analysis Test” 5. U.S. Utility Patent Application, “System and Method for On-Line Impulse Frequency Response Analysis”

MECHANISM OF COKING IN LOAD TAP CHANGERS A STATUS REPORT

Russ Crutcher Microlab Northwest

Dave Hanson TJH2B Analytical Services

Leo Savio ADAPT Corp.

Abstract The objective of this EPRI funded project is to postulate mechanisms of coking in LTCs; to design controlled studies for verifying proposed coking mechanisms; to formulate a plan that addresses maintenance modes, monitoring techniques, operating practices, and design specifications for reducing coke formation. The project team has examined and tested coked samples from operating LTCs to discern the sequence of physical and chemical events that occurred to them. The team evaluated LTC components from the component surface through the deposit to the surface of the coke formation. They evaluated insulating fluid samples for starting materials, intermediate species, and products that are converted to coke at the surface of coke formations. The team also examined interactions between the fluid components and the coking surface and has determined that there are several causes of coking, the chemical processes being the most complex. In all cases, the coking mechanism started with an organic or inorganic non-conducting coating of the contacts. Many precursors to the formation of these films have been identified. To refine its hypotheses, experiments are being conducted using small test cells containing oil and copper parts. INTRODUCTION The purpose of the test cells was to create an environment where changes that occur in an LTC over long periods of time can be created in a much shorter period of time. This accelerated aging would be the result of using higher temperatures, higher concentrations of challenging agents, or creating continuous conditions that simulated exposure over several cycles in a functioning LTC. The number of parameters that may be important to the formations of stable carbon films is large, and so is the number of parameters that can be controlled in the test cells. Testing all of these parameters independently would be prohibitive in cost and time. The approach taken was to test "regions of interest" within the multiple dimensional matrix space defined by these variables. Figure 1 illustrates how this was done. A sub-field of three

dimensions was selected for each set of tests. In theory three parameters are being tested while all of the other variables are being held constant. Each of the three variables is changed within the selected space to determine how these three variables interact. In point of fact all of the other variables are active and are fluctuation over some small range. As long as those uncontrolled fluctuations are negligible the effects of the three variables of interest can be determined and test results can be confidently repeated. If the tests cannot be reproduced within acceptable limits then one or more of the uncontrolled variables is changing in a way that is significant and must be actively controlled. The materials in the test cell are monitored chemically and physically to characterize the changes that occur as a function of the parameters being tested. The measurements made on these materials include the list of parameters below. Output Parameters: Temperature by Position on the Conductor Change in the Weight of Conductor (film weight) Chemical Properties of Film by Position Crystallographic Properties of Film by Position Mechanical Properties of Film by Position Dissolved Gases in the Oil Sulfur Content of the Oil Poly-Aromatics Present in the Oil Furans Present in the Oil Mechanical Properties of Oil Temperature of the Oil

Y-axis

Z-axis

X-

Figure 1 A three Dimensional Sub-Space of an N-Dimensional Study Space

Test Conditions In this phase of the study we wanted to evaluate selective sub-populations of the multiple dimensional matrix space (time, temperature, cycling, furans, arcing, sulfur, aromatics, oxygenated alphatics, etc.) and generate a model challenge that could be used to evaluate control methodologies. Our test configuration also allowed us to test a range of temperatures in each run. The copper test rod was four inches long and was bored out to contain a two-inch heater unit that was inserted into the rod. The surface temperature of the copper rod was monitored at three locations using thermocouples that were fixed to the rod with wire. The heater element was set for a specific temperature through a feedback loop between the controller and a thermocouple in the heater itself. The lower thermocouple location on the rod in the test cell was below the heater position. This location was heated by conduction through the copper by the heater and was cooled by conduction by the oil in the tank that flowed over its surface. Under these conditions the gradient range of temperatures increased with increasing test temperature because of the natural increase in convection flow at the higher temperature. Results Three basic types of films were created during these tests: inorganic, organic, and carbonized films. The inorganic films were of two basic chemical types, copper oxides and copper sulfides. The inorganic films were often complex and dynamic through a series of tests. Low temperatures tended to create mixed oxide and sulfide films. At higher temperatures the film on the surface of the rod tended to be an oxide film and the sulfides had moved into the carbon rich part of the film. These inorganic films ranged from less than 40 nanometers to over a micrometer in thickness. Often these films generated characteristic interference colors that could be used to determine film thickness. The organic films were waxy, non-crystalline polymers. All of them had a distinct orange to redorange color. They all exhibited shrinkage cracks. These materials were highly coordinated carbon ring structures with elevated carbon-carbon bonding, as was indicated by the high refractive index (>1.80) shown by the film. The carbonized films were black films indicating higher carbon coordination . Some of these films were very tenaciously bonded to the surface of the copper. The tenaciously bonded carbon films seemed to be always associated with second and third order inorganic films on the surface of the copper. Thermal Effects In the temperature range from 60 to 90 Celsius the copper was a very effective scavenger of oxygen and sulfur from the system. These films were invariably present at some level after the test and were not present at detected levels prior to the test runs. The films on the surface of this rod are about 40 nanometers thick and include sulfide species as well as oxides of copper.

Organic films didn’t form when new oil was used but when used oil or oil with 10 parts per million added naphthalene was used an orange film would form over the heater. This film was effectively removed by a tapelift and could be easily flaked from the surface. At these temperatures the bonds between the film and the underlying copper were weak. Flakes of the film were removed and tested for their refractive index. The refractive index of the film was found to exceed 1.80. Refractive indices in this range indicate a significant loss of hydrogen and the formation of a higher ratio of carbon-to-carbon bonds in the film. In the temperature range of 130 to 200 Celsius complex films were always present, even when new Diala was used. There were often two distinct films, one inorganic film a few tenths of a micrometer in thickness on the surface of the copper and an organic film above the inorganic film, which often exceeded ten micrometers in thickness. Sulfides were always present but their location varied. Test rods that had relatively thin or poorly formed carbon films still had sulfides in the inorganic layer of film. As the film carbonized and became thicker the sulfur tended to leave the inorganic film on the surface of the copper and enter the carbonized film. Often sulfur could not be detected in the inorganic film but was strongly indicated in the carbon film. This duplicates what has been seen in real world samples but the mechanisms involved have not been determined. The organic films covered the test rods as long as the temperature over the entire rod exceeded 130 Celsius. Copper apparently reacts readily with the oil at these temperatures to generate these films. Below 130 degrees Celsius new Diala was relatively unaffected over the forty-eight hour duration of the standard test. Over the heater element the organic films carbonized. The temperature at this location on the rods always exceeded 160 Celsius. The lower thermocouple recorded the coolest temperatures on the rod and during this run that temperature well exceeded 130 degrees Celsius. The carbon films have not been characterized by x-ray diffraction at this time but they will be so characterized as part of this phase of the project. Effects of Additives The effects of additives was briefly addressed above but will be considered in more detail here. There were three basic types of additives considered: inorganic (primarily sulfur), furans, and aromatics. The effects of sulfur were of interest because it was invariably present in the films covering failed contacts in the "real" world. Furans were of interest because they are present in used transformer oils and to varying degrees in new oils exposed in some LTC's believed to be free of cellulose based products. The aromatic compounds are common in all oil exposed to the LTC environment. Their source includes arcing events and reactions occurring on the surface of heated copper or silver. The amount of these chemical challenges added to the cells was invariably small in terms of the total volume of oil in the cells but it was significant in terms of the total weigh of film formed during the various tests. Each of the additives exhibited some characteristic patterns supper-imposed on the thermal patterns discussed above. The additives will be considered individually below. Inorganic Additives

The principle inorganic reactants are oxygen and sulfur. The oxygen content was not controlled in these runs and was considered to be elevated. Sulfur was added as elemental sulfur in solution. At 30 part per million sulfur the effects were significant. The film that formed at a temperature of from 170 to 200 Celsius was uniformly black and powdery. It had little cohesion and the tapelift simply removed the top layer of debris, which left the rod still looking black. The copper beneath the film had a layer of copper oxide nearly a 100 nanometers thick in places. None of the other films examined during this phase of the testing have lacked the cohesion of the films created by adding elemental sulfur. It was also interesting that all of the sulfur was located in the black film and that the inorganic film on the surface of the copper was all copper oxide. Furan Additives Furfuryl alcohol and 2-furaldahyde have been added as representative furan compounds. They have been added at 0.25, 1, and 25 parts per million. Figure 7 shows the test rods after exposure to 0.25 and 1 part per million furfuryl alcohol. The total area covered with carbon film seemed to be greater with the lower concentration of furfuryl alcohol though the film thickness was less over the heater. This series of tests was run without changing out the Diala between tests. After the 0.25 ppm run a new rod and an additional 40 microliters of furfuryl alcohol was added. The decrease in the area of coverage may have been the result of the depletion of some other component in the new oil. This is being investigated. The film over the heater is similar in structure for both rods though the higher concentration of furan is associated with thicker films in each of the three distinct layers present.

Effects of Aromatic Compounds One of the challenges used consistently in these tests has been the "used" oil. This material is oil from a variety of sources, both transformers and LTC's. It was shown early in the study to result in films forming on heated copper long before new oil showed any effect. For that reason we have repeatedly returned to this material as a target film generator for a procedural or pure compound additive that could be used as a consistent reference material. The "used" oil contains a variety of aromatic compounds and includes furans. The effects of adding a pure aromatic (naphthalene) in the low temperature runs was impressive in that it was similar to the behavior of the "used" oil. No duplicate comparison as yet been made in the 130 to 200 temperature range. That test will be made as part of this phase of the study. The used oil tested in this range produces a multiple layer film in some ways similar to that seen with the furans. This may be the result of the presence of furans in this oil. That will be evaluated in future tests conducted with arc-conditioned oil and with pure compound additives. The "used" oil generated films were consistently dark in color and were thick by comparison to the furan generated films. The film was brittle and cohesive. It lifted easily with the tapelift but did leave section of film still adhering to the copper oxide layer. The oxide layer is relatively

thick, about 400 nanometers and a shinny blue film is evident. The duller black film exhibits layering. DISCUSION We have been able to generate carbon based films on copper heated in new Diala oil, with and without additives. By modifying temperature, time, and additives in the test cells we have created films with averaged growth rates spanning four orders of magnitude, from 0.1 milligrams per hour to 100 milligrams per hour. The films generated exhibit a variety of structures, some similar, but not identical to the films seen in “real world” samples. We are now at the point of more precisely reproducing results so that we can develop a reference procedure against which to compare alternative materials. This will require more replicate test runs and more analyses of both the films and the oil used in the tests. We have seen differences in the properties of films using different additives. Here again repeat tests will help determine if these differences are characteristic of the additive or were a coincidence associated with other variables. If the association of specific variables and film characteristics can be verified then reference films associated with specific conditions in the equipment could be identified. All of the above are required to develop a “Control Film”. A film that can be reproduced with optimal reliability so that it can be used to evaluate more subtle differences between the behavior of different oils or different configurations of materials.

CONCLUSIONS One of the surprises in this study is how easy it is to form organic films on copper. Films form at lower temperatures than previously expected and with a variety of contaminants that are commonly encountered in oil with age. There are a variety of different structures in the films created under different conditions but there are also consistent patterns. At this stage in the project the knowledge of carbon based films in LTC's has already been greatly increased. The different types of carbon deposits seen in LTC's have been described and the four basic mechanisms responsible have been identified. The unique identification of chemically stable films and their role in the failure of contacts better defined a previously unknown cause of failure. A procedure for evaluating these films has been developed and is already aiding in the diagnosis of contact failures.

A number of tasks still need to be accomplished as part of this phase of the study and are in progress. The project is on schedule and our original goals still seem to be within the scope of this project. This is still a work in progress but it has already generated information that has been applied to working systems to prolong their useful life. More cost benefits will be realized as the project continues.

On-Line DGA During Controlled Energization of a 1000 MVA Autotransformer Exhibiting Partial Discharge Thomas Waters, Sr. Chemist Dan Morgan, Dir., Engineering Serveron Corporation Hillsboro, Oregon, USA

Abstract: In June 2001 National Grid conducted a controlled back energization test of a large power transformer that was known to have partial discharge problems. The test was conducted using a variable voltage mobile generator connected via a transformer to the 33kv tertiary winding of the transformer. The transformer was a 1000MVA 400/275/33kV unit belonging to Scottish Power at Neilston substation west of Glasgow, Scotland. The transformer had been taken out of service in 1997 following a Buchholz gas alarm. Partial discharge activity had been indicated by raised levels of acetylene and hydrogen in the oil. The recent experiment was arranged to test a wide variety of partial discharge detection technologies from several different companies and universities worldwide. With sponsorship and support by EPRI, on-line dissolved gas monitoring was included in the repertoire of equipment used to observe the characteristics of the transformer. Increases in the dissolved gases due to discharges inside the transformer were detected by the dissolved gas monitoring systems within a matter of a few hours after the discharges occurred. Additionally, the on-line gas monitoring allowed for better understanding of the nature of the discharges. Oil samples were taken by National Grid personnel before, during and after the test for comparative purposes. This paper presents the results of the experiment from a dissolved gas perspective.

Introduction: Several years ago National Grid conceived of an experiment to re-energize a “failed” but intact transformer that had a known history of partial discharges for the purpose of evaluating the ability of state-of-the-art monitoring equipment to detect transformer faults as they occurred. The experiment was conducted from June 25-28 2001 at the Scottish Power - Neilston substation west of Glasgow, Scotland. The monitoring equipment under test was primarily focused on real-time detection of the partial discharges that were produced when the transformer was re-energized. Scientists and engineers from several different companies and universities worldwide were present with a wide array of equipment to detect the partial discharges. In addition, the Serveron TrueGas – Oil Phase instrument was invited and sponsored by EPRI to monitor the transformer during the experiment from an on-line dissolved gas detection perspective. An overview of the facility, transformer, and monitoring equipment used is provided to familiarize the reader with the experiment and equipment at the site. Details of the experiment are discussed below followed by data from the experiment, data analysis, and conclusions. The Facility: The Neilston substation is located approximately fifteen miles west of Glasgow, Scotland, in the Renfrewshire council area, a short distance outside the village of Paisley. The station is owned and operated by Scottish Power. National Grid was in charge of managing the overall experiment including setting-up and conducting the experiment, recording data and timing of the energizations, compiling the data on the National Grid web site, and for site cleanup once the experiment was over. The transformer and its immediate surrounding area were fenced off to isolate access to portions of the transformer yard not directly involved with the experiment.

The Transformer: The test transformer was a Hackbridge & Hewittic, 1000 MVA, 400/275/33 kV core-form three phase auto-transformer with ONAN/OFAF cooling system, built in January 1972, and commissioned at the Neilston facility in September 1977.1 The transformer has a free-breathing conservator with a 15,000-imperial gallon oil capacity. There are two circulation pumps on the cooling system, each capable of delivering 826 imperial gallons/minute1, allowing a theoretical volumetric turn-over of the transformer oil approximately 6 times per hour. The transformer does not have oil flow diverted through the core. Therefore, any gases generated during the experiment in the oil inside the core would only mix with the main circulated body of oil through diffusion or convection if enough heat was generated during the experiment. The transformer had a history of gassing behavior associated with partial discharges and was rebuilt in 1996-1997. The unit was energized and very shortly thereafter began to generate gases again. In June 1997, the transformer tripped the Buchholz relay. When the Buchholz was sampled after the failure, the hydrogen was found to be 318885 ppm and the acetylene was at 822 ppm. The methane was reported at 800 ppm. This data confirmed the large gassing event within the transformer. The asset was subsequently taken out of service and the main electrical connections to the transformer were removed. Only minimal maintenance was performed on the transformer between the failure and the experiment. National Grid - Dissolved Gas Analysis (DGA) ppm After Transformer Failure in 1997 Date Sampled

Oxygen

Nitrogen

Hydrogen

Carbon Monoxide

Carbon Dioxide

Methane

Ethylene

Ethane

Acetylene

Sampled From

6/23/97

13208

61433

97

462

3322

19

1

2

52

Main Tank

6/25/97

72706

628810

318885

45779

5213

800

96

21

822

Buchholz

Table 1 – DGA Results from Test Transformer After 1997 Failure 2

1 – Ratings as recorded on the transformer Name Plate. 2. – DGA data provided by National Grid

TrueGas Installation: The TrueGas analyzer is an on-line gas chromatograph that can automatically analyze eight fault gases in dielectric insulating oil from transformers (H2, O2, CH4, CO, CO2, C2H4, C2H6, and C2H2). The analyzer was mounted on the east side of the transformer, approximately 1.5 feet from the main tank of the second phase of the transformer. Oil was sampled from the transformer’s cooling loop return to the main tank. Approximately 23 feet of ¼-inch O.D. stainless steel tubing was used to route oil to the TrueGas analyzer. Oil was returned from the

analyzer to the transformer through the transformer’s bottom drain valve. Approximately 19 feet of ¼-inch O.D. stainless steel tubing was plumbed for this connection. An isolation transformer was connected to the TrueGas analyzer so that it could be powered by 110VAC. The TrueGas instrument was installed four days prior to the start of the experiment to allow the readings to establish a baseline trend on the gases dissolved in the oil. Installation and operation were in accordance with the manufacturer’s published guidelines. The Experiment: Applied Voltage: The test was conducted over a three day period using a variable voltage mobile generator (Aggreko) connected via a transformer to the 33kv tertiary winding of the test transformer. The test transformer had no load applied for the experiment but the applied excitation voltage was taken to approximately 440kV phase to phase. The actual voltage applied to the transformer during the experiment was measured at a bushing tap, which had been previously calibrated from a 10kV calibrated external source. Additional voltage measurements were made at the generator terminals during the tests to act as an indication. At the highest voltages at 50Hz (the normal operating frequency) magnetizing current harmonics caused significant voltage waveform distortion, this was somewhat reduced by using a 60Hz excitation frequency.[1] Conversions from percent generator voltage to applied voltage for both phase-to-neutral and phase-to-phase are listed in Table 2.

Generator Voltage (%)

Voltage phase-neutral Voltage (kV) (kV)

70

160

277

85 90

213 225

337 356

95

238

376

100

250

396

105

263

416

110

275

436

phase-phase

Table 2 – Conversions from % Generator Voltage to Applied Voltage3

Oil Circulation: Because the dissolved gas monitoring equipment was sampling oil from the transformer cooling loop return, it was essential that the oil be circulated through the transformer to assure that the oil being sampled for on-line DGA was as representative as possible. Without circulation, the oil being sampled for both the on-line monitoring and the DGA oil samples would not have been representative of oil exposed to the discharges within the transformer. The cooling loop circulation pumps were turned ON for testing purposes three days prior to the experiment and oil was circulated for approximately one hour. Audible circulation pump noises were observed during this initial test. The pumps were again activated at the beginning of the first day of the experiment. No other pump noises or signs of “cavitation” were observed for the remainder of the experiment. The pumps continued to recirculate the oil in the transformer until the middle of the second day of the experiment when the pumps were turned off for thirty five minutes to validate whether the pumps were causing additional noise in the partial discharge detection equipment. The pumps were turned back on again and remained on until the experiment was completed. Experiment Timed Events: A complete timeline of events was recorded by National Grid during the experiment and contains exact times and durations of the energizations by the percentage generator voltage applied, the frequencies, and oil turbo pump activities. The experiment was conducted over the course of three days with a total of ten distinct energization experiments, totaling nine hours and nineteen minutes of excitation of the transformer. Six hours and thirty five minutes of this time were at the higher applied voltages where discharge signals were detected. The actual duration of the observed partial discharges was relatively brief. Discharges within the transformer were confirmed by virtually all of the partial discharge detection groups when the transformer was operating at 85% of generator voltage which was calculated to be 337 kV (phase to phase). ______________________________________________________________________________ 3 – Table generated at the experiment by National Grid.

As asset damage was a concern, the voltages applied initially were less than those applied at the end of the experiment. Additionally, the frequency of the applied voltage was also varied over the course of the experiment; 55.7Hz, 50Hz, and 60 Hz were used. A summary of energizing of the transformer is provided in Graph 1.

Percent Generator Voltage versus Time by Frequency % V at 50 Hz

%V at 55.7 Hz

%V at 60 Hz

% Generator Voltage

120 100 80 60 40 20 0 6/25/01 12:00

6/25/01 18:00

6/26/01 0:00

6/26/01 6:00

6/26/01 12:00

6/26/01 18:00

6/27/01 0:00

6/27/01 6:00

6/27/01 12:00

6/27/01 18:00

Date / Time

Graph 1 – Energization of the Transformer during the Experiment4

Dissolved Gas Analysis: National Grid took oil samples before and during the experiment for DGA analysis. The results from these DGA are listed in Table 3. A baseline sample was taken from the main tank of the transformer on June 6, 2001. The circulation pumps had not been activated on the transformer for some an unknown period of time when this sample was taken, but it was surmised that it had been at least several months and possibly years since the transformer oil had been mixed. National Grid - Dissolv ed Gas A naly sis (DGA ) p p m Before and Du ring th e Ex p erim en t Date S am p led

Carbo n Oxy g e n Nitro g e n Hy dro g e n M o no xide

Carbo n Dio xide

S ample d M e thane Ethy le ne Ethane

A c e ty le ne

Fro m

6/6/01

20795

60385

15

58

513

6

5

2

8.5

M ain Tank

6/27/01

11520

36048

66

98

322

9

7

1

20.2

M ain Tank

6/27/01

14946

39045

62

92

341

9

7

1

18.1

Buc hho lz

6/28/01

14596

32989

82

69

298

9

8

1

24.1

M ain Tank

Table 3 – Dissolved Gas Analysis Results for Oil Samples Taken Before and During the Experiment4

National Grid personnel sampled the oil from the transformer by draining oil into a 500 ml glass bottle through a plastic tube. An initial 100 ml of oil was used to rinse the container and was then discarded. The sample bottle was filled and sealed with a plastic pressure septum cap. An air bubble of approximately 5-10 ml was left in the container to allow for thermal expansion of the

oil. The bottle was inserted into a black plastic bag so that the sample would not be exposed to direct sunlight. The sample was then appropriately tagged and shipped for analysis. TrueGas Results: The TrueGas analyzer is capable of analyzing eight gases dissolved in dielectric insulating oil up to once every four hours. At twelve hour sampling, the analyzer established a baseline reading for all detected gases during the three day period prior to the start of the experiment. The analyzer was set for an automatic four hour sampling interval for the duration of the experiment. A total of twenty five analyses were performed on the transformer oil by the TrueGas instrument in six and one half days of operation. It was observed that there was a very slight reduction of sample gas pressure during the four hour sampling interval so the analyzer was instructed to not perform a sample at noon each day to allow the instrument to maintain full equilibrium. It was subsequently discovered that the instrument was not consistently achieving the necessary flow rate of oil due to automatic oil pressure shut-downs. Upon investigation, it was observed that the conservator on this particular style of transformer is elevated approximately 35-40 feet above the base of the transformer. The positioning of the conservator in conjunction with the pressure generated by the cooling loop circulation pumps created oil pressure spikes above 32 psig and a background head pressure of 19 psig. This caused the oil pump within the instrument to stop periodically to protect the system from over4 – Data provided by National Grid

pressure damage. Once the pressure dropped below the set point, the oil pump would automatically restart. Due to the high number of observed pressure spikes throughout the experiment, the number of these restarts was abundant which effectively slowed the oil flow rate through the instrument and in turn lengthened the time it took to reach equilibrium. The TrueGas analyzer calibration is field verified against a certified eight gas mixed standard, typically mounted inside the analyzer enclosure. A verification analysis was performed on the instrument once/day during the experiment. These results are listed in Table 4. The accuracy of all of the verification analyses were within the performance standard for the instrument. Concentration Verifications of the TrueGas A naly zer (p p m ) Ca rb o n

Ca rb o n

Hy dr o g e n

Oxy g e n

Me tha ne

Mo no xide

Dio xide

Etha ne

Ethy le ne

A c e ty le ne

Date / Tim e

[H2]

[O 2]

[CH4]

(CO)

[CO 2]

[C 2H6]

[C 2H4]

[C 2H2]

S tandard Co nc e ntratio ns

5000

1000

500

750

3000

500

250

100

6/25/2001 9:54 6/26/2001 17:49 6/28/2001 8:59

4988 4971 4985

973 955 964

501 494 496

764 758 769

3021 3008 3017

499 493 492

256 254 254

104 105 106

A v erage

4981

964

497

764

3015

495

255

105

S tandard De v iatio n (ppm)

8.92

8.95

3.91

5.65

6.72

3.87

1.18

1.20

% diffe re nc e o f S tandard

0.38

3.59

0.57

-1.84

-0.51

1.05

-1.97

-4.80

Table 4 – Verification Concentration Data from TrueGas Before and During the Experiment

The data presented for TrueGas is raw data. All data filters within the instrument were disabled to allow for the highest resolution of the data. The oxygen concentrations detected in the transformer oil by the TrueGas analyzer were above the measurement capabilities of the instrument. From DGA data listed in Table 3, the oxygen concentrations were above 14500 ppm and are considered the most representative data for oxygen. The methane concentrations detected in the transformer oil were below the detection limit of the analyzer. The DGA the concentrations were below 10 ppm and showed no changes during the experiment as reported by the DGA data. The most significant results from the experimental data came from the analysis of the hydrogen and acetylene data (Refer to Graph 2). The hydrogen [H2] started at a baseline of 40 ppm and rose to 42 ppm after the first day of experiments. This slight increase is likely due to the brief time of excitation where discharging occurred on the first day. The second day of the experiment generated a 19% increase to 50 ppm. After the third day, an additional 54% increase was observed with a concentration of 77 ppm. The hydrogen concentration increase for the experiment, from 40 ppm to 77 ppm, represents a total increase of 93% or 37 ppm.

Graph 2 –Hydrogen, Acetylene, and Ethylene TrueGas and DGA Results for Experiment

The hydrogen data from the TrueGas correlated well with the DGA results. Comparing the DGA results of 66 ppm from the second day of the experiment to the TrueGas reading of 63 ppm, a difference of 4.8% was observed. The DGA for hydrogen from the Buchholz sample valve was 62 ppm, a difference of 1 ppm. The last reading from the TrueGas instrument for experiment was at 8:00 AM on the third day and read 77ppm. The DGA sample was taken at 10:00 AM from the main tank and reported 82 ppm of hydrogen a difference of 6.5%, but the TrueGas hydrogen being observed was increasing at a rate of 1.1 ppm/hour. Following the trend observed from the previous three analyses, the instrument would have reported a number very nearly that of the DGA on the next analysis if the run had been made. The acetylene [C2H2] started at a baseline of 12 ppm and rose to 13.5 ppm after the first day of experiments. The second day of the experiment generated a 33% increase to 18 ppm of acetylene. After the third day, an additional 54% increase was observed with a concentration of 34 ppm. The increase for the experiment, from 12 ppm to 34 ppm of acetylene, represents a total increase of 183% or 22 ppm. The acetylene from TrueGas agrees well with DGA results also. The second day DGA results of 18.1 ppm and 20.2 ppm were lower than the 26 ppm from TrueGas, but agreed well with the 18.2 ppm from the previous TrueGas reading. The final DGA result was lower than the final TrueGas reading too, but again looked better when looking at the TrueGas trending. Ethylene [C2H4] was detected just above the lower detection limit for the instrument. The TrueGas analyzer reported a baseline of 5 ppm and a rise to 8 ppm after the third day. This was confirmed by all of the DGA samples which reported to within 1 ppm of the TrueGas readings.

As the changes in ethylene concentrations over the course of the experiment were only 2 ppm, little can be inferred from this data except that the gas generation observed was most likely not due to hot metal faults. Ethane [C2H6] was detected just below the detection capabilities of the analyzer for most of the experiment. The chromatographic peak was identified by the analyzer, but the concentrations increased to above the detection limit of 5 ppm only for the last analysis of the experiment giving only a single valid data point from the TrueGas for ethane. The DGA results were lower at 1-2 ppm, but confirmed the very low concentrations. The variability in carbon monoxide can be seen in Graph 3. The TrueGas instrument reported carbon monoxide starting at a baseline of 140 ppm, but dropped to 123 ppm after the transformer cooling pumps were turned on the first day of the experiments. For the second and third day of the experiments, the analyzer reported increases in CO from the baseline to 140 ppm the second day and to 149 ppm the third day. Both times, the concentrations recovered to the baseline of approximately 123 ppm. The increase is confirmed by a slight observed increase in the oil temperature recorded at the TrueGas analyzer, but the temperature increase is not enough alone to account for the increase. It was observed during the experiment that the top of the main tank was quite warm especially over the second phase of the transformer. D is charge Activity vers us T rueGas and D GA D is s olved Gas Concentrations of Car bon Monoxide % Ge ne ra to r Vo lta g e

Oil Te m p (°C)

Ca rbon Mo noxide (CO)

DGA (CO) 160

120

100

120

80

100 80

60

60 40

Concentration (ppm)

% Generator Voltage and Temperature (°C)

140

40 20

20

0 6/23/010:00

0 6/24/010:00

6/25/010:00

6/26/010:00

6/27/010:00

6/28/010:00

6/29/010:00

Date / Tim e

Graph 3 – Carbon Monoxide from TrueGas and DGA compared to Discharge Activity and TrueGas Oil Temperature.

The TrueGas readings and DGA concentrations reported for CO were lower just before and during the experiment than were recorded after the fault in 1997. It is theorized that as the transformer cooled, some of the gases adsorbed into the cellulose insulation. When the insulation was warmed due the energization and discharges of the experiment, some of the adsorbed gases were liberated. When the system cooled over night, the gases appear to have re-absorbed leveling to the baseline concentrations once again as shown by the TrueGas data. This behavior has been witnessed before and is predicted by H. Kan’s 1995 analysis [2]. The DGA results for carbon monoxide from the samples taken during the experiment do not correlate as well as those from hydrogen and the C2 hydrocarbons. The results from the DGA varied by 29 ppm or and were between 34 to 48% lower than the corresponding TrueGas measurements.

The reported data for the carbon dioxide concentrations were similar to those of carbon monoxide as shown in Graph 4. The same cyclic nature is evident in the carbon dioxide data as well, but not to the degree of the carbon monoxide. The changes were less than 5% of the total concentration, unlike carbon monoxide which varied by as much as 20% of total concentration. As with carbon monoxide, the DGA results for carbon dioxide did not correlate as well as the other gases. Here too, the samples were between 22 to 32 % lower than the corresponding TrueGas analysis data. D is charge Activity vers us T rueGas and D GA D is s olved Gas Concentrations for Carbon D ioxide

% Generator Voltage

Carbon Dioxide [CO2]

DGA [CO2]

140

500

130

% Generator Voltage

110

400

100 90

350

80 300

70 60

250

50 40 6/23/01 0:00

6/24/01 0:00

6/25/01 0:00

6/26/01 0:00

6/27/01 0:00

6/28/01 0:00

200 6/29/01 0:00

Date / Tim e

Graph 4 – Carbon Dioxide from TrueGas and DGA compared to Discharge Activity

Concentration (ppm)

450

120

Observations from the Data: From the rate of gas generation, it does not appear that any discharges were produced at the lower voltage tests (< 85% of the generator voltage applied) conducted on the first day of the experiment. When comparing the times of the excitations with the responses of the TrueGas analyzer, the generated gases took between 4 and 8 hours to show up in the TrueGas data with the cooling loop circulation pumps operating. After the second day of experiments the hydrogen rose 8 ppm. At the same time, the reported acetylene increased by 4.5 ppm for a ratio of 1.8:1 , [H2]:[C2H2]. After the third day of testing the hydrogen rose an additional 27 ppm. The acetylene during the same period rose 16 ppm for a ratio of 1.7:1, [H2]:[C2H2]. For the total experiment, the hydrogen rose 37 ppm and the acetylene rose 22 ppm for a ratio of 1.7:1. A hydrogen- acetylene ratio this low indicates that the discharges and subsequent generation of gasses were at very high temperatures, most likely exceeding 700°C. This infers that the induced discharges within the transformer were from arcing events as opposed to lower temperature partial discharges that would be characterized by a high generation of hydrogen with small increases in methane and traces of acetylene [3]. Conclusions: 1) The TrueGas on-line monitoring system can detect the occurrence of discharges within circulated transformer oil in a few hours. 2) The nature of the gassing during the experiment suggests that the predominant fault mechanism was higher temperature arcing as opposed to partial discharges. 3) From the activity of the carbon monoxide and carbon dioxide, not all of the changes in gas concentrations were due to generation. There is evidence that supports the claim that these two gases were most likely desorbed and reabsorbed into the solid insulation of the transformer as temperature varied.

Acknowledgments: Serveron Corporation would like to thank EPRI for their support and sponsorship that made attendance and participation in this experiment possible and for inclusion of this paper at the EPRI Substation Equipment Diagnostics Conference X. Additionally, Serveron Corporation would like to thank National Grid for their outstanding cooperation in helping to setup the TrueGas analyzer and in managing the experiment in a safe and effective manner.

References: 1. Partial Discharge Experiment web site maintained by National Grid, http://www.partialdischarge.co.uk 2. H. Kan, T. Miyamoto, “Proposal for an Improvement in Transformer Diagnostics Using Dissolved Gas Analysis (DGA),” IEEE Electrical Insulation Magazine, Vol.11, No.6, November/December 1995. 3. IEEE C57.104-1991 “Guide to Interpretation of Gases Generated in Oil-Immersed Transformers” (ANSI)

Profitability Assessment of Transformer On-Line Monitoring and Periodic Monitoring Jacques Aubin, André Bourgault General Electric Canada Inc.

Claude Rajotte, Pierre Gervais Hydro-Quebec

Summary Throughout the electrical industry, the need is recognized to support any significant capital investment with a cost/benefit analysis. This rule also applies to monitoring of power transformers. However, no recognized economic model is yet available to assess the profitability of transformer monitoring. The method proposed in this paper is based on the economic model developed by IEEE for circuit breaker on-line monitoring. Key parameters are identified and analyzed in two scenarios: with and without monitoring. Numerical examples are presented for various transformer operating conditions. 1 Introduction The cost/benefit analysis of transformer monitoring is a difficult undertaking. Calculating the cost is relatively straightforward, taking account of equipment purchase, installation, training and maintenance costs. The benefits however are more difficult to assess as the evaluation relies partly on factors dictated by experience. This exercise involves the probability of an event actually occurring, compounded by the positive or negative impact of such an event. Some events lend themselves to an economic analysis: • Reduced inspection and maintenance costs • Reduced failure-related repair or replacement costs • Improved real-time transformer loading capability • Deferred upgrade capital costs due to load growth • Deferred replacement capital costs due to equipment age or condition There are also a number of other benefits that are tangible but cannot be quantified easily: • Enhanced system reliability and availability with fewer unplanned outages • Improved planning for scheduled outages by using remote equipment condition assessment to avoid additional outages • Enhanced financial results with performance-based regulation • Optimized design and operating practices • Reduced commissioning costs • Increased equipment life • Enhanced personnel safety • Improved environmental safety • Retained knowledge of most skilled staff (expert system) • Improved system-wide access to key knowledge using Web tools These aspects are discussed at length in the draft IEEE Guide for Application of Monitoring to Liquid Immersed Transformers and Components(1).

Even for those items that lend themselves to an economic analysis, some preliminary economical values have to be established for conditions that are not usually quantified: • Failure rate reduction to be expected after introduction of monitoring • Cost of operating with a degraded transmission system • Value of energy not generated • Value of contractual energy not delivered In spite of these difficulties, an attempt is made below to quantify some of these aspects in order to allow the “business case approach” to be applied whenever a project for on-line monitoring or additional periodic monitoring is to be implemented. In the following sections, different operation scenarios are considered. In all cases, comparison is made between the situation with and without monitoring as recommended in the draft IEEE Guide for the Selection of Monitoring for Power Circuit Breaker(2). Many details suggested in this document have been left out in this presentation but should be considered in a complete analysis. 2 Inspection and Maintenance Costs In many cases, additional monitoring can reduce the frequency of manual inspections, especially for transformers with a tap changer. Direct time savings are achieved during disassembly, manual inspection, reassembly and reporting. Moreover, depending whether the monitoring is applied to a single transformer or to a whole substation, there can be additional savings in travel time and vehicle expenses. On large power transformers, it is common to have a monthly visit for an overview of the substation, completed by a more detailed visual inspection with oil sampling for DGA once a year. Since there is presently no monitoring system that can completely replace the visual inspection, it is felt that annual inspections will remain necessary in the foreseeable future and therefore no savings are accounted for in the example of calculation below. A more elaborate inspection, with outage, is carried out typically every 6 years and includes cooling system and protective equipment verification along with a number of electrical tests. This inspection calls for about a few tens hours-person. It is not clear if monitoring could allow postponing of these activities and therefore no savings are accounted for in the example of calculation below. The situation can be quite different for a utility undergoing more frequent inspections. A third group of maintenance activities deals specifically with tap changers. This preventive maintenance typically takes place after 40,000 to 100,000 operations depending on the type of unit, with a maximum duration between inspections. It is plausible that some savings could be achieved on this activity if a dependable tap changer monitoring is implemented. For example, assuming that a monitoring system would allow extending from 4 to 6 years the duration between a maintenance activity costing $5,000, the annualized cost reduction could be as presented in Table 1. Table 1 - Preventive Maintenance Benefit From On-Line Monitoring

Preventive Maintenance Costs Preventive LTC maintenance cost without monitoring Number of years between maintenance without monitoring Preventive LTC maintenance cost with monitoring Number of years between maintenance with monitoring Subtotal: Annualized preventive maintenance costs Annual benefit from preventive maintenance reduction

Annualized Costs With Without Input Values Monitoring Monitoring $5,000 $1,250 4 $5,000 $833 6 $1,250

$833 $417

Table 1 shows that improved tap changer monitoring only results in a saving of $417 per year for the LTC preventive maintenance aspect. It can be concluded that contrary to some other substation equipment such as circuit breakers, the savings on inspection and maintenance of power transformer is rather modest. 3 Failure Resolution Cost The most recognized benefit of early detection of incipient faults is the major savings that can be achieved on repair costs(3). In this regard, the purpose of an on-line monitoring system is to prevent major (or catastrophic) failures and convert them into failures that will be repaired at a reduced cost during a planned outage. In order to assess the savings on the failure resolution cost, preliminary values such as expected reduction of failure rate must be estimated. 3.1 Impact of Monitoring on Failure Rate In order to evaluate the economic gain derived from deployment of additional periodic or on-line monitoring, the full picture of fault evolution and detection must be laid down. It must be recognized that some of the slow-evolving faults will be detected anyway by existing devices such as gas accumulation relay and winding temperature indicator. Current monitoring practices such as annual gas sampling for DGA analysis will also detect the slow-evolving faults. At the other end of the spectrum there are some failures that are instantaneous by nature and not susceptible of early detection whatever monitoring system is installed. In between are the fastevolving faults that cannot be detected correctly by existing means but could be detected by suitable monitoring. It is on these faults that benefits can be achieved in the failure resolution cost. The breakdown of failure rate in different categories is depicted in Figure 1 with an example of values. This breakdown of failure type has been proposed before(4) and it is found useful in the calculation of the potential benefits to be derived from transformer monitoring. •

The transformer failure rate is the figure applicable for major faults on the transformer (or family of transformers being considered). Major fault is usually understood as fault leading to a removal of the unit from its base or a site intervention on the active parts. Typical failure rates rest between 0.5% for very reliable equipment and 3% for the very problematic. A figure of 1% is used in the example below.

Transformer failure rate: 1% 0.01

Detectable with existing means: 30% (Bucholz, DGA)

Not detectable with existing means: 70% 0.007

0.003

Detected with on-line monitoring: 60%

Not detected: 40% (including instantaneous failures) 0.0028

0.0042

Catastrophic failure avoided: 10% 0.00042

Non-catastrophic failure avoided: 90% 0.00378

Figure 1 - Breakdown of Failure Probability •

•

•

•

•

Some of these faults are detectable with existing means. Existing devices such as gas accumulation relays, top oil temperature indicators and hot-spot simulators can provide warning to developing faults. Periodic inspection and oil sampling for DGA analysis is known to be an efficient monitoring procedure. The proportion of faults that can be detected by these means can be estimated to be 30%. The remaining faults are not detectable with existing means. This portion is therefore estimated here to be 70% of all major faults. Some of these faults may display precursory signs but too tenuous to be detected by existing devices. This is the type of fault that is targeted by on-line monitoring. It would be unrealistic to expect a detection efficiency of 100%. Depending on the type of monitoring system deployed, some faults can still go undetected or develop at a rate too fast to allow for proper alarming and orderly removal from service. In the example below the detection efficiency is estimated to be 60%. The faults not detected include those that are instantaneous by nature, for instance an insulation breakdown following a lightning surge or severe short-circuit. Moreover, some components such as bushing shields are prone to sporadic failures that occur without any warning. Non-catastrophic failures are defined here as the major failures contained within the transformer tank, while catastrophic failures are those involving tank rupture and

eventually fire with collateral damage to nearby equipment. It is useful to distinguish between these two failure modes since the economic consequences are quite different. It is assumed here that catastrophic failures account for 10% of total failures. The benefit of monitoring is not to prevent a fault from occurring but to prevent it from developing into a major failure without any warning to the transformer operator. Therefore the gain from reduction in major faults will be somewhat attenuated by an increase in minor faults or predictive maintenance actions. The two scenarios (with and without monitoring) compare as follows: Without Monitoring (Present Conditions)

With Monitoring

Major failure occurring without advance warning

0.007

0.0028

Major failure prevented by early detection of fault

0.003

0.0072

Total

0.01

0.01

The faults currently detected at an early stage (by existing means) can be subtracted from both sides and the cost will be compared between the following two situations:

Major failure occurring without advance warning

Without Monitoring (Present Conditions)

With Monitoring

0.007

0.0028

0

0.0042

Additional predictive repair work

To complete the benefit evaluation in regard to failure resolution, values must be assigned to the average repair cost under each scenario: Repair cost for major failure without any advance warning Replacement cost and collateral damage in case of catastrophic failure Repair cost for early detection

$1,500,000 $5,000,000 $200,000

3.2 Failure Resolution Analysis for Individual On-Line Monitoring The cost/benefit analysis can be made either for an individual transformer or for a family of transformers when the deployment of monitoring systems is targeted for a number of units with a similar failure rate. In the case of centralized monitoring equipment or portable equipment that is to be used for periodic monitoring, the whole family must be considered since the cost of some equipment will be spread over a number of units. All costs are annualized for the case of a transformer without monitoring and compared to the case with monitoring (refer to Table 2). The benefits resulting from avoidance of major and catastrophic failures need to be evaluated separately because the economic consequences are very different. Table 2 - Failure Resolution Benefit From On-Line Monitoring

Failure Resolution Costs Current rate of not detectable failures Expected monitoring system efficiency Proportion of failures that are catastrophic Major Failure Resolution Cost Repair cost for major failure Repair cost for early detection Catastrophic Failure Resolution Cost Replacement and collateral damage Repair cost for early detection Annualized failure resolution cost Monitoring System Cost Monitoring system cost Expected life of monitoring system (years) Installation cost Annual support for monitoring system ($/year) Annual monitoring cost Subtotal: Cost for failure resolution and on-line monitoring Annual benefit from failure resolution cost

Annualized Costs With Without Input Values Monitoring Monitoring 0.007 60% 10% $1,500,000 $200,000

$9,450

$3,780 $756

$5,000,000 $200,000

$3,500

$1,400 $84 $6,020

$12,950

$2,000

$40,000 20 $5,000 500 $0 $12,950

$250 $500 $2,750 $8,770 $4,180

3.3 Failure Resolution Analysis for Periodic Monitoring of Tap Changer Some monitoring systems are inclusive and cover most components of the transformer. In other cases, the monitoring system is dedicated to a single component such as a bushing or tap changer. We may consider for instance the implementation of additional periodic monitoring for tap changers on a group of 50 units that are worth $2,000,000 each. This is a periodic monitoring method where the acoustic signature of a tap changer is compared with previous records or with signatures from similar equipment. To analyze the economic benefits of the implementation of this method, the following values need to be established (with numerical example): • Number of transformers in the target group 50 • Failure rate applicable for this group 0.5% • Proportion of transformer failures attributed to tap changer 50% • Failure reduction expected from implementation of new monitoring 50% • Average cost of a tap changer failure (50% of rewinding cost) $375,000 • Cost of monitoring equipment $20,000 • Annual labor cost per transformer for acoustic monitoring $100 The prospective benefits resulting from this monitoring are assessed in Table 3.

Table 3 - Benefit From Periodic Acoustic Monitoring of Tap Changers Annualized Costs for the Target Population Without With Input Values Monitoring Monitoring Cost for LTC Failure Resolution Number of units in the target population 50 Transformer failure rate 0.5% Failures attributed to tap changers 50% Failure reduction expected 50% Average repair cost of LTC failure (per unit) $375,000 Cost of LTC failure resolution $46,875 $23,438 Annualized failure resolution cost $46,875 $23,438 Cost for Periodic Monitoring of LTC Number of units in the target population Test equipment cost Depreciation period (years) Annual monitoring labor cost (per unit) Annual monitoring cost for 50 units Subtotal: Cost for failure resolution and periodic LTC monitoring Annual benefit from periodic LTC monitoring (for 50 units)

50 $20,000 10 $100

$2,000

$0 $46,875

$5,000 $7,000 $30,438 $16,438

3.4 Cost of Lost Generation Failure of a generating station unit (GSU) can have major economic impacts when the loss of a transformer leads to generation deficit and purchase of replacement power. This can be the case for a hydroelectric generating station in the high water season where the loss of a transformer results in water shedding. It can also be the case for any GSU at peak load season where a loss of generation leads to purchase of replacement power. Although the cost attached to the loss of a generating station unit can be seen from various point of view, the following values (with numerical example) have to be taken into account: • Current GSU failure rate 0.007 • Expected efficiency of the monitoring system 60% • Power not generated in case of GSU major failure 20 MVA • Duration of outage 30 days • Cost of replacement energy CRE The annualized cost of loss generation without monitoring can be estimated as 0.007 ∗ 20MVA ∗ (30∗24)h ∗ CRE $/MWh and in this example, this cost could be reduced by 60% with addition of monitoring.

3.5 Cost of Contractual Power Not Delivered Most transmission networks have sufficient transformer capacity to sustain a transformer failure without interruption of service to the user. However, the network will then be operating in a degraded condition and might not be able to sustain a second incident without interruption of service to the end user, thus leading in some cases to penalties for unfulfilled contracts. For instance in Australia, the legislator has established the “Value to the community of lost load (Voll)” to be $10,000/MWh(5). The cost of contractual energy not supplied is not related so much to the normal selling price but rather to the value established by regulating agencies as a penalty for not delivering power. Considerable work has been done around the world on estimating the customer interruption costs. Data have been collected by the CIGRE Task Force 38.06.01 from many countries. As indicated in their report (6), the cost of power interruption is very sensitive to the outage duration and the type of load; it can exceed $10,000/MWh for a few hours interruption. There are some costs associated with the operation of a degraded system. The cost of risk associated with a second contingency is the product of the probability of a second contingency occurring during the transformer outage, multiplied by the economic consequence of that event. The second contingency might occur as a result of a minor failure on the backup transformer or on any associated equipment that would prevent the backup transformer from carrying its duty. Minor failures are the ones to be considered because the probability of occurrence is typically an order of magnitude higher than the major failure rate. Still, a minor failure can last from a few hours to a few days and can be of significant economic consequence if the power supply is interrupted. The cost of the risk associated with a second contingency is calculated as follows (with typical number provided as example): •

Probability of a second contingency occurring during the main transformer outage: it is the product of probability of major failure on the main transformer (0.007) by the probability of minor failure on the backup transformer and associated equipment (0.15), taking account of the outage duration on the main transformer (150 days); this outage duration can be much shorter if a spare unit is available at site or in the transformer storage area. Second contingency probability = 0.007 ∗ 0.15 ∗ 150/365 = 0.000432

•

Cost of interrupting a delivery of 80 MVA for 10 hours at a penalty price of $10,000/MWhour Cost = 80 MVA ∗ 10 hours ∗ $10,000/hour/MWh = $8.0M

•

Cost of risk is the product of probability of the event multiplied by the consequences of this event Cost of second contingency risk = 0.000432 ∗ $8.0M = $3,456/year

The annual benefit on contractual power not delivered due to on-line monitoring is evaluated in Table 4. It is assumed that an early detection of the problem would always allow for planned outage without interruption of contractual power delivery. Table 4 - Cost for Contractual Power Not Delivered Annualized Costs Without With Input Values Monitoring Monitoring Penalties for Unfulfilled Contracts Current rate of not detectable failures 0.007 60% Expected monitoring system efficiency Duration of outage on main transformer (days) 150 Probability of minor failure on backup transformer 0.15 Duration of minor failure on backup transformer (hours) 10 Power not supplied (MVA) 80 Value of energy not supplied ($/MWh) 10,000 Subtotal: Cost for contractual power not delivered $3,452 $1,381 Annual benefit on contractual power not delivered $2,071

4 Reinforcement of Overloading Capability Power transformers have inherently some margin of overloading capability. The rated capacity of a unit is basically the load level that will result in internal temperatures not exceeding the limits set forth by standard producing bodies such as the IEEE and IEC. The application of loads in excess of the nameplate rating involves a degree of risk and accelerated aging. These effects are discussed at length in the IEEE and IEC loading guides(7-8) and can be summarized as follows: • For short-term transformer failures, the main risk is the reduction of dielectric strength due to the release of gas bubbles in regions of high electrical stress. The probability of occurrence of these bubbles is closely related to the winding insulation hot-spot temperature and the moisture content of the insulating paper. • Under overloading conditions, some components such as LTC contacts and bushing connections may develop high temperatures leading to thermal runaway. High temperatures may also occur in structures when the stray magnetic field increases beyond the saturation point of magnetic shields. • For long-duration overloads, the main consequence is the thermal aging of the solid insulation. With time, the cellulose chains undergo a depolymerization process, thus reducing the average length of the cellulose chains and consequently reducing the mechanical strength of the paper. This paper deterioration is a function of time, temperature, moisture content and oxygen content. This effect is irreversible and forms the basis for transformer life duration. 4.1 Effect of On-Line Monitoring on Overloading Capability It is generally recognized that the risks associated with overloading can be significantly reduced if the transformer conditions are closely monitored throughout the overload period(9-11). To quantify this benefit, the additional loading margin provided by monitoring needs to be stated.

Field experience with transformer overloading is still limited, but it can be conservatively assumed that if the parameters mentioned below are properly monitored, the transformer can carry an extra 10% loading with the same degree of confidence that would exist in operation without monitoring. A comprehensive monitoring system, comprising dissolved gas-in-oil and moisture sensors, provides a major support to the operator when the transformer faces overload conditions: • Continuous monitoring of hydrogen and carbon monoxide provides dependable detection of excessive overheating of leads, shields, structural parts, contacts or bolted joints. It will also alert the operator if excessive circulating current is taking place in the core or in the tank, leading to arcing or overheating of contact points. • The moisture sensor measures the relative moisture content of the insulating oil. Over time, the on-line monitoring system can use this information to calculate the moisture content of the insulating paper, thus allowing estimation of the bubble inception temperature. This temperature is compared to the hot-spot temperature to indicate the safety margin prevailing during the overload. The moisture content of the insulating paper is also an important factor to take into account when calculating the insulation’s loss of life. • Although thermal loss of life of the winding insulation cannot be prevented, on-line monitoring can provide a dynamic evaluation of insulation degradation. The relative loss of life can then be converted into cost if some hypotheses are set regarding normal life duration and transformer cost. The cost attributed to loss of life needs to be subtracted from the apparent benefits achieved from transmitting this extra load. The loss of life cost can also be useful when calculating the cost of transmitting this additional amount of energy. 4.2 Cost/Benefit Evaluation for Overloading For the specific case of transformer overloading, the following parameters need to be considered in order to quantify the benefits: • The additional loading margin is the extra loading allowance made available by on-line monitoring by reducing the inherent risks associated with high load levels. As an example, it could be estimated that a transformer can be loaded up to 110% without special monitoring and that on-line monitoring will allow loads up to 120% with the same degree of confidence. • The transformer replacement cost (including engineering and installation) is needed to calculate the value of the loss of life incurred during the overload. In the forthcoming example, this value is assumed to be $2,000,000. • The transformer normal life duration is a conventional reference for continuous duty under normal ambient temperature and rated operating conditions. The IEEE Loading Guide(7) indicates a number of benchmarks that can be used. In the calculation below, a remaining degree of polymerization of 200 is selected as the end-of-life criterion, leading to a normal life duration of 150,000 hours. This value applies for a rated winding hot-spot temperature of 110 °C with minimum content of moisture and oxygen. • The loss of life is directly related to the winding hot-spot temperature that can be calculated from prospective load profile and ambient temperature. The aging acceleration

•

factor increases exponentially with the temperature. In the simplified example presented below, a continuous 110% loading leads to a hot-spot temperature of 120 °C and an aging acceleration factor of 2.7. For a 120% loading, the hot-spot temperature is assumed to be constant at 135 °C with an aging acceleration factor of 11. Therefore 100 hours at that temperature is equivalent to 1100 hours at the rated temperature of 110 °C. The economic value ascribable to the overload alone is obtained by subtracting from the total the aging caused by the normal load. The net loss of life can then be related to the normal life duration and the transformer cost to quantify the value of loss of life attributable to the overload. The profit margin on the extra energy delivered during a peak demand period can be very high. In the calculation below, it is assumed to be $80 per MWh.

Using this simplified assumption along with steady load and steady ambient temperature, the benefits of on-line monitoring have been assessed in Table 5. Table 5 - Transformer Overloading Benefit From On-Line Monitoring Annualized Costs Without With Input Values Monitoring Monitoring Benefit from Overloading Capacity Extra loading without monitoring (MW) 10 Extra loading with monitoring (MW) 20 Duration of overloading (hours) 100 Probability of overload occurrence 0.20 Profit margin of delivered energy ($/MWh) (80) Transformer normal life duration (hours) 150,000 Replacement cost of transformer $2,000,000 Aging acceleration factor at 110% load 2.7 Aging acceleration factor at 120% load 11.0 Gross profit on extra energy delivered ($16,000) ($32,000) Value of additional loss of life $453 $2,667 Subtotal: Cost for overloading ($15,547) ($29,333) Annual benefit from overloading with monitoring $13,787

5 Deferring Transformer Replacement Whenever a transformer is considered for replacement, a comprehensive condition assessment is usually carried out. This would imply a number of electrical and chemical tests, some of them requiring an outage, leading to a diagnostic on the insulation condition. In this context, the economic value of on-line monitoring is difficult to assess, as the decision must also take into account other considerations such as capital availability, delivery time for new units and the perception that the insulation condition could start to degrade in the near future. Considering the large financial consequence of deferring transformer replacement, it is however worthwhile to attempt an evaluation of the economic contribution that could be expected from on-line monitoring.

In some cases the transformer replacement is considered to eliminate concern raised by either the age of the unit or a known defective condition. In other cases the consideration for replacement arises from normal load growth, the transformer eventually reaching its loading limit. This loading limit is usually applicable to an emergency condition that would prevail during an outage on the parallel transformer, since substations are normally provided with some degree of redundancy. As a result, transformers usually spend most of their life at a load below their nameplate rating(10). During that period, the failure rate is practically independent of the load level. As the load increases, the transformer will be requested to carry higher loading, and as paper aging progresses the insulation may degrade; this is the beginning of the worn-out stage. The failure rate can then be expected to increase depending on various factors such as the moisture content, the tightness of the windings, and the temperature of leads connections and tap changer contacts. The transformer would normally be removed from service if the failure rate were believed to rise beyond the acceptable level. At this stage, condition monitoring becomes critical to maintain an acceptable level of reliability. By providing early detection of incipient faults, on-line monitoring reduces the risk of unexpected failures, raises the reliability to an acceptable level, and allows for postponing the transformer replacement. If half of the major failures can be avoided and converted to minor failures, reliability is enhanced and the transformer can be allowed to serve for a few additional years before an unacceptable probability of failure is reached (see Figure 2).

Failure rate %

Without monitoring

Infant mortality

Steady state

With monitoring

Worn out Acceptable level of failure rate Replacement deferral

0

10

20

30 Years

40

50

60

Figure 2 - Effect of Monitoring on Transformer Life Duration The benefit from deferred replacement is directly proportional to the current interest rate and the capital cost of a new unit. For example, deferring replacement of a $2,000,000 transformer at a current interest rate of 6% is worth $120,000 annually.

As the transformer population grows older, the replacement scenario will become even more relevant. Additional efforts are needed to quantify more exactly the economic relation between closer condition monitoring and the serviceability of a unit near the end of its useful life. 6 Conclusions Economic analysis of transformer monitoring is a difficult undertaking since some of the critical numbers have to be estimated from experience or derived from “educated guess”. However such an exercise cannot be avoided any more as there is increasing pressure to rationalize capital investments and direct the available funds toward the most profitable areas. A detailed calculation method has been proposed by the IEEE for the economic analysis of circuit breaker monitoring. A similar approach has been developed in this study, although only the most significant numbers are presented in the calculation sheets. The main cost and benefit items are reviewed and presented in a comparison of two scenarios: with and without monitoring system. This approach appears to provide useful results, which will gain even more credibility when the performance of the proposed monitoring system is well established. The proposed calculation method lent itself to sensitivity analysis to test the criticality of the most debatable numbers. It has the merit of forcing the user to address the truly significant parameters. Results are presented in terms of annual savings for a single unit or a family of transformers. Conversion can be made to more convenient economic indicators such as Net Present Value or Payback Period. Models for the economic analysis of transformer monitoring are still in development by the IEEE Transformer Committee and the CIGRE Transformer Study Committee. It is hoped that this contribution can be helpful in the development of these economic tools much needed in the industry. References 1. IEEE P C57.XX Draft 9, March 2001, “IEEE Guide for Application of Monitoring to Liquid Immersed Transformers and Components” 2. IEEE C37.10.1-XX Draft 7, “Guide for the Selection of Monitoring for Power Circuit Breaker” 3. T. Krieg, “Management of Transformers – Tools and Techniques for New Age Asset Management”, CIGRE Colloquium, June 2001, Dublin, Ireland 4. Brekeenbridge, Harrisson, Lapworth, McKenzie & White, “Economic Issues Concerning Loss Capitalisation and Monitoring”, CIGRE Colloquium, June 2001, Dublin, Ireland 5. P. Austin, “Transformer Economic Issues in Australia & New Zealand”, CIGRE Colloquium, June 2001, Dublin, Ireland 6. CIGRE Task Force 38.06.01, “Methods to Consider Customer Interruption Costs in Power System Analysis”, CIGRE Technical Publication 191, August 2001 7. IEEE C57.91-1995, “IEEE Guide for Loading Mineral-Oil-Immersed Transformers” 8. IEC 354, “Loading Guide for Oil-Immersed Power Transformers”

9. W.J. Bergman, “Equipment Monitoring Selection as Part of Substation Automation” Panel Session, IEEE Winter Power Meeting, New York, 1999 10. T. Molinski, “Minimizing the Life Cycle Cost of Power Transformers”, CIGRE Colloquium, June 2001, Dublin, Ireland 11. R. Farquharson, “Technology Solution for Improving the Performance Reliability of Substations and T&D Networks”, Energy Association of Pennsylvania, Hershey, PA, September 2001

Monitoring of Power Transformers - PSE&G Experience

Larry Johnson PSE&G, USA Oleg Roizman IntellPower, Australia Valery Davydov Monash University, Australia

Abstract This paper describes field experience gained at PSE&G with monitoring various substation transformers. The paper will focus on many new and innovative transformer monitoring and diagnostic techniques being developed under an EPRI sponsored project and implemented at an EPRI member utility, PSE&G. The goal of this project is to employ a cost-effective method to capitalize on existing and future microprocessor devices utilized in the substation environment to develop a robust monitoring system. Other objectives of this project were: to· apply and evaluate the Monash moisture assessment method on selected power transformers of different types, sizes, and winding arrangements; to establish the merits and identify limitations of on-line moisture assessment for the selected transformers; and to assess the performance and diagnostic capability of commercial moisture-in-oil sensors. Special attention will be given to the effects of oil age on accuracy of moisture determination.

Introduction Utilities throughout the world have experienced the damaging effects of moisture in power transformers. Some of these damaging effects result in long-term outages, prolonged filtering processes and even catastrophic failures. Many studies have been performed to define the consequences of water contamination. Many of these studies define adverse transformer impacts ranging from decreased dielectric strength, loss of life, tank and core corrosion to vapor bubble evolution. Other studies have defined acceptable levels of moisture, mechanisms of moisture ingress, the dynamics and interactions between the oil, winding insulation, temperature, load, etc., as well as safe operating limits for “wet transformers”. These studies, theories, and models typically utilize and require manual methods of sampling, recording, analyzing, and evaluating test data for theory verification. These manual processes require the Utility to spend an inordinate amount of time and resources for this process. Moisture in oil/paper insulating systems continues to plague the Utility Industry and continues to have an adverse affect on the long-term health of transformers. To protect this expensive asset it was determined the Monash algorithm should be tested under real life conditions utilizing new tools, technologies and systems in hopes of eliminating the resource intensive methods currently utilized. Under EPRI sponsorship, Monash University and PSE&G have partnered to utilize the dynamic behavior of moisture in oil/paper insulation systems from previous studies and apply the findings to an on-line system. The intent of utilizing the results from this new concept of 'water-in-paper activity' is to develop an on-line system that eventually employs fuzzy-logic based techniques to assess the water-in-paper activity in transformer insulation systems.

Description of Transformers PSE&G presently owns and operates Transmission and Distribution facilities across the state of New Jersey. Transformers included in this study were chosen to measure and determine the effectiveness,

accuracy and the benefits associated with an on-line monitoring system to determine the wetness or dryness of different size transformers in PSE&G’s inventory. Based on the final results of this project a determination may be made of which type, class, and size transformer would be best suited for monitoring systems. A decision was made to select five transformers for the study. The decision was based on age, availability of design data, load, location, existing infrastructure, and ease of installation. It was determined that a 4MVA (26kV – 4kV) transformer from Rahway Substation, a 27/36/45 MVA (138kV – 13.8kV) core form transformer from Devils Brook Substation, a 24/32/40 MVA (230kV – 13.8kV) shell form transformer from Somerville Substation, a 180/240/300 MVA (230kV – 13.8 kV) core form transformer from Athenia Substation, and a 550MVA (230kV – 138kV) core form transformer from Fairlawn Substation would be part of this study. The Rahway transformer is in a station that provides power to approximately seven thousand residential customers in the city of Rahway. This transformer was built in the early 1900’s and is moderately loaded. This station design is typical for a station that provides service to many older cities and has shown little load growth, will not be replaced or changed in the near future and still provides reliable service to the customers. Keeping these assets healthy intelligently and economically is vital to both the system and customer. The Devils Brook, Somerville and Athenia transformers generally speaking are part of a more modern substation design serving loads that include residential, light commercial and industrial customers. The transformers range in age from two to thirty five years old. With new growth in the areas these stations are beginning to reach their capacity and during peak periods overloading can occur. Maintaining these assets effectively is critical because of loading concerns, contingency planning, and financial ramifications. The Fairlawn unit is a very big and expensive asset that is a critical path for load flow in PSE&G territory. Knowledge of the moisture content and the ramifications of the dynamics of this moisture will provide PSE&G with better loading, energizing and maintenance decision tools.

Architecture The ability to validate, perfect, automate, and implement the Monash algorithm cost effectively on a variety of power transformers in the PSE&G system was the overall goal of this study. To accomplish this task the appropriate sensors, software and hardware needed to be defined and tested. The first task was to define the platform that would automatically log, store, and transfer data files in a standard, universal format. The archived data needed to be available automatically or upon request and in a format that requires little manipulation from end users. Based on previous experience implementing a standardized platform for microprocessor relays Softstuf’s Substation Data Concentrator (SDC) was selected to be the platform of our substation information gathering system. The SDC is currently deployed for microprocessor relay integration and has historically provided the System Protection Group within PSE&G with a means to obtain rapid fault information from many different microprocessor relays minutes after an unwanted event has occurred. The SDC captures fault data from microprocessor relays and places the record on the corporate LAN for viewing. In parallel fault location, fault type, date and time are sent to the Operations Department for restoration purposes. Utilizing this familiar substation hardened device the SDC was augmented and modified to be the platform for sensor integration, data logging, automatic data transfer, intelligent alarming and future system monitoring. Softstuf’s integration package and user-friendly tool set provided for easy transition, development and inclusion of the new monitoring demands. It was determined the SDC needed to sample defined variables at defined intervals simultaneously and then archive this data in a comma delimited or .csv format. This format was selected to ensure consistent data point timing and to fulfill the universal analysis tool criteria. After the platform, data format, and data sampling rates were defined the log files or .csv files were sent to Monash University electronically for evaluation according to a specified schedule. Once the integration platform was defined sensor definition, selection, and optimal sensor placement were considered high priority. To facilitate this process the quantities or variables required for this project were first defined. The quantities to be measured were top and bottom oil temperature, relative saturation, 1-3 ambient temperature, and transformer load. Based upon previous research the Vaisala HMP 228 sensor for transformer temperature and moisture in oil sensing was recommended. The Vaisala HMP 228 sensor provided accuracy to 2% moisture and was easy to integrate in the SDC. The Vaisala HMP 228 uses a standard ASCII protocol, provided the necessary documentation to assist in automation, and could

be configured to communicate via RS-232 or RS-485. The Vaisala sensor has installation options that allow the user to select different probe and cable lengths, uses standard thread sizes for valve installation and can be mounted in small areas with ease. This sensor is measuring top and bottom oil temperature in Rahway Substation. The remaining sensors for this study included a suitable ambient temperature sensor, an accurate, user friendly, easy to install meter to measure transformer load, and a transformer temperature sensor that was accurate, rugged and had a digital output for remote communication. The next device to be investigated and evaluated was the ambient temperature sensor. After comparing different temperature sensors for accuracy, dependability, functionality and cost the Advanced Power Technologies sensor was chosen. This device was selected based on its accuracy, ease of use, value added features, ease of system integration, and the manufacturer’s willingness to work with the end user. To measure top oil temperature in the bigger units of this study the Weschler Advantage sensor was selected. At the time of this initiative Weschler Instruments had developed a device that could provide this project with a suitable digital replacement for our old analog transformer temperature sensors. It included remote communication capabilities, calculated winding hot spot temperature, and could measure load at a reasonable cost. The Advantage was installed easily and communicated with the SDC using RS-485 over copper. This device requires more testing and evaluation before further deployment system wide. To provide this study with accurate transformer loading data and capture the opportunity to investigate and evaluate new substation grade microprocessor meters various meters were evaluated. After investigating and evaluating a number of solutions and devices it was determined that the easiest, most accurate, and cost effective way to gather load data was to employ and install the SATEC PM-172E digital meter. The SATEC meter was found to be user friendly, provided very accurate readings under extreme conditions, has numerous communication protocols to choose from, and has peripherals to complement it product line and make integration easier. SATEC also had the willingness to accommodate the end user. Other concerns related to communication methods and mediums, data logging, sampling rate, and data formats will be discussed. After all devices were defined the system was integrated and tested as a single entity. During this evaluation period device/system weaknesses were exposed and rectified.

Installations Installation objectives sought were to develop a standard system configuration that focused on optimizing sensors, meters, and existing infrastructure. There was also the opportunity to capitalize on value added endeavors to enhance asset utilization, predictive maintenance systems, and system restoration along with study participation. PSE&G used this opportunity to make accuracy comparisons between devices that advertise multi-functionality with single function meters. Each station was equipped generically with the same family of devices, which led at times to generic solutions. It was decided to install the SDC inside each control house and bring the information into this device via RS-232 or RS-485 depending on the location of the sensor. Sensors that required physical mounting on the transformer included the Viasala HMP 228, and the Weschler Advantage. Quantities that could be obtained from devices inside the control house were transformer load and ambient temperature. All transformers involved in this study were situated at various distances away from the control house and all in the outside environment. The shortest distance was approximately 100 feet from the control house and the longest distance was approximately 400 feet from the control house. This situation alone allowed PSE&G to develop an understanding of the costs, benefits, and considerations one must take to effectively monitor a device in an existing substation yard. Fiber optics appeared to be the logical choice in stations that required long runs and no conduit or ducts. The determination was made that if digging was required fiber would provide the most reliable means of communication and would be immune to electric fields within the substation. The cost is in the digging and laying conduit thus negating the cost concerns associated with fiber. In the two locations where conduit or duct existed copper was pulled with ease and success. Both communication mediums require converters or modems to communicate with the SDC and have all performed admirably. Conclusions concerning the best communication configuration should be defined on a case-by-case basis and in PSE&G’s case increased service time is necessary before determining the best components and configuration. Sensor installation and placement provided the greatest challenge. Every transformer required the addition of valves and had mounting, installation and wiring concerns. In some instances special steel

plates needed to be fabricated to allow the mounting of valves on the transformer. Some transformers required longer sensor probes, and longer probe wire. All transformers require outages and some required special switching to make the installation possible. To ensure proper sensor placement and to validate the Monash algorithm Karl Fischer tests were to be performed by the PSE&G Maplewood Testing Services facility. This portion of the initiative still remains an open item and will not be completed until more analysis is performed. Many devices exist in each substation that measure and record transformer load. Depending on the age and design of the station there were as many as three different meters that monitored transformer load. In most cases graphic ammeters, chart recorders and/or analog meters were the installation of choice when the station was built. Some newer stations have metering devices with serial ports that allow easy integration but are found to be somewhat inaccurate and unreliable. Therefore to provide the necessary accuracy and reliability a replacement meter for substation use was evaluated. The SATEC PM-172E meter was installed and utilized for this study and future substation replacement projects. Inside all station control houses the preferred communication method was via RS-232 on copper because of the short distances and ease of installation. At Rahway substation metering quantities were gathered from the SATEC meter mounted in an outside transformer cubicle and integrated with the SDC utilizing fiber and RS-232. The ambient temperature sensor selected had multiple sensor inputs and programmable output contacts that were utilized for other purposes and provided accurate temperature reading over a wide range. This device communicated with the SDC using an ASCII format via RS-232.

Data Analysis The following transformer parameters were requested and supplied along with on-line recorded data: transformer i.d., serial no, manufacturer, year of manufacture, years in operation, rating, voltage, core/shell type, cooling type, oil type, oil capacity, oil preservation, mass of paper, cooling elements, rated winding temperature rise, rated losses, winding conductor, weight of core and coils, weight of tank and fittings. In the following sections we present results of moisture assessment for each studied transformer.

Rahway Substation Figure 1 a, b, and c shows load, temperature and RS variations during two weeks in June 2001 recorded at Rahway transformer. Temperature at a top oil level varies within 40 – 60 °C, which results in the RS change from 30 to 40 %. Such high RS values indicated that the transformer is in a very “WET” condition. Analysis of response from the second moisture sensor installed at the bottom oil level supported this prediction for the transformer state in terms of moisture content. Determination of active water content of solid insulation was performed with the help of TMM (transformer moisture monitor) and explained in more detail in the further sections. It was determined that water concentration in paper can be as high as 4.9 %.

Athenia Substation Very low RS and calculated water content in oil (WCO) was observed during the entire period of monitoring. Comparison between calculated WCO and that measured by KF is presented in Figure 2 c. The following observations were made:

Laboratory analysis by KF resulted in a wide range of values from 10 to 30 ppm (Figure 2 c), which are considerably different from each other, although, theoretically, the WCO trend should parallel a temperature profile. This observation raises the question of consistency of WCO determination by the standard KF method. The lowest values of WCO by KF method are much higher than those predicted by the model based on temperature and RS. This raises the question of sensor location. In the natural oil flow cooling system (ONAN) the moisture desorbed from the paper exists within the vicinity of the winding and does not move far into the radiator unit. That may explain the

difference between absolute values of water content in the main transformer tank and the radiator. Another explanation, which can be given to support the above observation, is the fact that part of moisture can get condensed in the cooler and therefore the sensor does not detect the whole amount of moisture leaving the main tank. Devils Brook and Fairlawn Substations Both Devils Brook and Fairlawn transformers have very low water content in oil/paper insulation. Devils Brook unit has N2 oil preservation system and is relatively new. Thus its expected moisture content is low. Fairlawn autotransformer is 18 years old and has a conservator tank/bladder preservation system, which has proved to be very reliable, preventing the moisture from getting in contact with the oil. Forced oil flow makes the choice of moisture sensor location a reasonably easy task. A radiator return loop was suggested for the moisture probe placement.

600 500 400 300 200 100 IA(3/AM P S )

IB (3/AM P S )

IC(3/AM P S )

0 21-Jun-01

23-Jun-01

25-Jun-01

27-Jun-01

29-Jun-01

01-Jul-01

03-Jul-01

29-Jun-01

01-Jul-01

03-Jul-01

29-Jun-01

01-Jul-01

03-Jul-01

a) 70 60 50 40 30 20 10 Amb

T top

T bot

0 21-Jun-01

23-Jun-01

25-Jun-01

27-Jun-01

b) 80 70 60 50 40 30 20 10

R S top

R S bot

0 21-Jun-01

23-Jun-01

25-Jun-01

27-Jun-01

c) Figure 1 a) Load, b) Temperatures and c) RS trends for the period of June 21 – July 4, 2001 at Rahway S/Station

50 45 40 35

T, C

30 25 20 15 10 5 0 6-Sep

8-Sep

10-Sep

12-Sep

14-Sep

16-Sep Top rad.

18-Sep

20-Sep

22-Sep

24-Sep

26-Sep

18-Sep

20-Sep

22-Sep

24-Sep

26-Sep

18-Sep

20-Sep

22-Sep

24-Sep

26-Sep

Bottom rad.

a) 5 4.5 4 3.5

RS, %

3 2.5 2 1.5 1 0.5 0 6-Sep

8-Sep

10-Sep

12-Sep

14-Sep

16-Sep RS top

RS bot

b) 40

35

30

ppm

25

20

15

10

5

0 6-Sep

8-Sep

10-Sep

12-Sep

14-Sep

16-Sep

WCO top

WCO bot

KF

c) Figure 2 a) %RS, b) T and c) ppm trends for the period of 6 – 28 September, 2001 at Athenia S/S

Determination of Moisture in Oil There are two commonly used techniques to determine the moisture content in oil. The standard method is based on the well-known Karl Fischer titration reaction. This procedure has a few major drawbacks. Firstly, since moisture content varies with load, manual periodical sampling unlikely will coincide with worst-case scenario. Secondly, this method is prone to a high level of error. Therefore, providing realistic results requires impractically frequent sampling and time-consuming effort. Another method is based on the determination of water in oil content from continuous measurement of RS and temperature. This method requires reliable moisture sensor and knowledge of water in oil solubility characteristics.

Oil Solubility Tests Due to the aging of oil the water solubility characteristics may change with time. To avoid possible mistakes associated with incorrect values for water solubility, laboratory testing of oil was recommended. The laboratory testing of oil from Rahway transformer was performed at Monash University in Melbourne. These tests revealed that saturation limit for the tested oil is considerably different from that of brand new oil. Therefore, it was suggested that a number of tests be performed to determine oil solubility characteristics before any calculations of water in oil are made. Depending on the type of oil and the number of years in exploitation, the oil solubility changes. Neglecting this change can lead to the significant error in calculation of oil ppm value. 600

400

S(T), ppm

Aged Oil

200

New Oil

0 0

20

40

60

80

T, deg C

Figure 3 Water Solubility Curves for New and Aged Oil Figure 3 shows the water solubility curves for new and aged oil. It is apparent from Figure 3 that the aged oil can hold more water at the same temperature than the new oil. For instance, at T=60°C saturation limit S(T) for the new oil is 220 ppm while as for the aged oil S(T) can reach as high as 390 ppm. Data presented in Figure 3 illustrates the importance of measurement of the water solubility characteristic of oil S(T) for the validation of an installed moisture sensor and for the accurate assessment of water content in paper insulation. At the request of the Monash laboratory a few liters of oil were sampled from the aged Rahway transformer operating in the utility for more than 60 years. The oil was sampled from the bottom drain valve of the transformer. The transformer was fitted with two moisture sensors: one at the top of the tank and another at the bottom of the tank. Monash water solubility formulae of Figure 3 for aged oils have

been used for the calculation of water content of oil (WCO) in ppm. Also a few Karl Fischer (KF) measurements of WCO have been taken from a container with the oil on the day of its arrival at Monash. Figure 4 shows the moment of oil sampling at Rahway transformer made on September 28. The substantial temperature and RS transients were observed from the moisture sensor installed at a bottom oil level reflecting the moment of moisture take up on this day. Figure 5 shows the results of the KF measurement and predicted moisture in oil content calculated using corrected for aged oil formulae. The KF measurement performed at Monash laboratory shows a good agreement with the value of WCO calculated for aged oil. This result proves the accuracy of the formula derived for the oil of this particular transformer. 70

60

60

50

50

40

40

30

30

20

20

10

10

T, C

RS, %

70

0 0:00

2:00

4:00

6:00

8:00

10:00 T bot

12:00 T amb

14:00

16:00

18:00

20:00

22:00

0 0:00

RS

Figure 4 Temperature and Relative Saturation Transients During Oil Sampling at Rahway Station

90.00

80.00

70.00

60.00

ppm

50.00

40.00

30.00

20.00

10.00

0.00 0:00

2:00

4:00

6:00

8:00

10:00

12:00

ppm, aged oil

14:00

16:00

18:00

20:00

22:00

0:00

KF@Monash

Figure 5 Comparison of Water Content Determination from Measurements and by direct KF Method at Rahway Station: 28 September 2001 Another comparison between calculated and measured ppm values (this time at PSE&G) was made during October and November 2001. Figure 6 shows that a good agreement between predicted and measured WCO by KF was reached. 80.00

70.00

60.00

ppm

50.00

40.00

30.00

20.00

10.00

0.00 5-Oct

12-Oct

19-Oct ppm aged oil, bot

26-Oct

2-Nov

9-Nov

KF, ppm

Figure 6 Predicted and Measured Water Content at Rahway S/Station, October –November 2001

Determination of Moisture in Solid Insulation In this project we evaluated moisture in solid insulation using an earlier developed software application 4,5 called Transformer Moisture Monitor (TMM) . TMM acquires data from the Vaisala HMP228 transmitter and calculates the moisture content in oil in ppm units. Using an intelligent fuzzy logic decision-making algorithm, TMM predicts so called 'active' moisture content in solid insulation, using the Monash method for moisture determination in paper. Waterin-paper activity (Awp) is a critical factor that determines the state of dryness in power transformers. By measuring Awp, it is possible to predict the degree of water contamination in an oil/paper system. Calculation of water content in paper is directly based on the correctly measured Awp. In addition, waterin-paper activity can play a significant role in determining the breakdown risk of high voltage equipment and the possibility of bubble formation. The front panel of the TMM is shown in Figure 7.

Figure 7 Front Panel of TMM Presenting the Moisture Assessment Results for Rahway Transformer in July 2001 TMM uses the 'Green-Yellow-Red' (GYR) traffic light approach to identify and display the condition of the power transformer in terms of moisture content. The moisture state is indicated by the ‘Red’, ‘Yellow’ or’ Green’ lights, depending on the insulation conditions. The diagnostic power of the decision-making mechanism is strengthened by a confidence level, which takes the values of HIGH, MEDIUM or LOW. When the confidence level is HIGH the displayed value of WCP (%) becomes a true value reflecting the percentage of water in paper available for exchange with the surrounding oil. This screen presents the diagnosis results for the Rahway unit. The period of monitored data was about two weeks in July 2001. As can be seen from the screen the water content in oil changes in the range of 60 – 115 ppm. The transformer under study is in critical state in terms of moisture content of solid insulation. The “HIGH” confidence level in diagnostics is evidence of that displayed WCP value of 4.9 % is a true value of ‘active’ moisture content in paper. This allows us to conclude that the transformer is “WET” and requires further attention.

Future Work The development and successful application of Transformer Moisture Monitor in the PSE&G demonstrates the substantial technological progress that has been made in the area of moisture assessment of a transformer oil/paper insulation system. However, more work is required in order to explain more deeply the effect of moisture on the integrity of transformer insulation and on decision making on dry out process. The major future improvements in moisture assessment are summarized as follows:

Determination of the most suitable location for moisture sensors The accuracy of moisture sensor performance and the quality of diagnosis are dependent on such oil parameters as temperature, viscosity, flow rate etc. Therefore the location of a moisture probe is of high importance. To obtain more accurate and reliable results it is recommended that the probe be installed at a location in the top oil level as close as possible to the winding. The upper radiator pipe is suggested as the most appropriate position. However, other locations such as the upper filling valve, a valve on the bottom of the cooler and a drain valve are also acceptable. Placement of moisture sensor probe on the return cooler pipe is recommended only in case of a forced oil flow system. More research should be conducted to study the effect of moisture probe placement on moisture assessment in oil/paper system.

Continuous Improvement in the Accuracy and On-Line Calibration of Moisture Sensors Moisture sensors based on thin film technology are prone to a slight drifting over the period of exploitation. Therefore, identification of sensor malfunction, drift development and online calibration will become necessary features in the future. As more and more moisture sensors become available comparison between different units from different manufactures is recommended as part of the future work.

Incorporation of Moisture Algorithms into Dynamic Rating and Risk Assessment Programs Dynamic rating and risk management are two important issues when evaluating the load capability of an operating transformer. Moisture contributes to the aging of transformer insulation and should be taken into account in all calculations associated with life management of a transformer. Among the other issues to be considered for future work are the following:

• • • • •

Real-time on-line determination of water solubility in oil Development of the moisture assessment algorithms for shell form and nitrogen blanketed transformers. Further study into the effects of moisture on gas dynamics and dissolved gas analysis Investigation into the relationship of water-in-paper activity and bubbling conditions Implementation of WIP methods for field drying of power transformers

The work on these developments will significantly expand the diagnostic power of the current version of TMM and will add value to the existing transformer condition monitoring and diagnostic tools.

Conclusions This on-line monitoring project provided PSE&G with many valuable lessons for future installations and initiatives. A new understanding of the costs, benefits, and considerations concerning substation monitoring were realized. Installation costs, wiring concerns, and problems associated with installing and terminating fiber optic cable, copper, fiber modems, RS-485/RS-232 modems and power concerns were exposed and noted.

This project has examined, exposed, and answered many logistic and installation questions and has uncovered more uncertainty about the dynamics of moisture in oil. The ability to effectively utilize existing infrastructure, new technology, and universal tools has proven beneficial in many other company initiatives. Data gathered from the new meters, sensors, and various microprocessor devices has now become accepted and is an integral part of the Computer Maintenance Management System (CMMS), planning function and system restoration tool kit. It is planned to continue experimenting and participating in new initiatives that help PSE&G drive costs out of its system, help proactively manage its expensive assets, and provides a safe working environment for its employees and customers.

References 1

V.G. Davydov, O.M. Roizman and W.J. Bonwick, “Moisture Evaluation in Oil and Paper for Sealed Transformer Insulation System,” Proceedings of EPRI Substation Equipment Diagnostics Conference VII, New Orleans, 1999, 15 p.

2

U.S. Patent Application, "Apparatus and Method for Determining Water Content in a PaperOil Insulation System.” Inventors: O. Roizman, V. Davydov, W. Bonwick and S. Lindgren. Filing Date: September 14, 1999, Serial No. 09/395416.

3

V.G. Davydov, O.M. Roizman, P.J. DeKlerk and M. Thomas, “Field Experience with Moisture Assessment in Power Transformers,” Proceedings of EPRI Substation Equipment Diagnostics Conference VIII, New Orleans, 2000, p.131-144.

4

O. Roizman and V.G. Davydov, “Role of Moisture Assessment in Condition Monitoring and Diagnostics of Power Transformers”, Proceedings of EuroTechCon Conference, Birmingham, 2001.

5

Transformer Moisture Assessment Method: Phase 3: Testing of Nitrogen Blanketed Transformer Insulation and Development of Software Application, EPRI, Palo Alto, CA:2001. 1001939.

TOWARDS MAINTENANCE-FREE LOAD TAP CHANGER (LTC) II

Hans-Ulf Schellhase, Ed A. Hall, Robert G. Pollock Powertech Labs Inc. 12388-88th Ave Surrey, B.C. V3W 7R7

Abstract The Electrical Power Research Institute (EPRI) and B.C. Hydro, British Columbia (Canada) sponsored this study on improvement of maintenance for "on-load tap changers" (LTCs). This second phase of the research addresses, among others, the topics of transformer oil quality for LTCs and the evaluation of oil-immersed contact materials. Parallel to this work, Powertech Labs Inc. continued its study on the diagnostic value of an LTC database for preventive substation maintenance. The paper investigates the significant differences between the chemical breakdown and ageing of mineral-based insulating oil in transformers and the breakdown of oil in oil-filled switchgear (Topic I). This description is followed by a presentation of new findings, which addresses the development of high resistance in stationary reversing switch assemblies (Topic II). Keywords: OLTC, LTC – Monitoring – Maintenance, Oil-filled Switch Gear, Tarnish, Coke Formation, Reversing Switch, Load-carrying Stationary Switches

Introduction For more than seventy-five years, on-load tap changers (LTCs) for transformers assumed a major role in the reliable delivery of electrical power. On-load tap changers serve the regulation of voltage under load without interruption. There are resistor and reactor (inductor) types of LTCs. On-load tap changers are in fact the single biggest contributors to transformer outages, and are worldwide statistically responsible for about one-third of all failures.

Load tap-changers contain, as major mechanical components, oil-immersed switchgear of different designs. This sub-equipment may accommodate up to three sets of three different switches in one oil-filled, three-phase compartment. These are usually a reversing switch, a diverter switch (current transfer) and a tap selector; the diverter switch will usually dissipate the brunt of the energy associated with contact making and contact breaking due to arcing. In a normal operating load tap-changer the diverter switch is the most common source of fault gases and carbonaceous particulates. For the present time, and the foreseeable future, the operator of a substation has to assess the maintenance needs of the old, and not so old existing equipment. A move to predictive, condition-based maintenance is a diagnostic challenge. The different LTC makes and models are matched by as many different oil-volumes of switch tank(s), open breather versus closed type, power dissipation characteristics etc. The power dissipation characteristics usually result from a combination of arc, glow discharge and thermal loss, which will induce chemical changes in the oil, and cause the formation of combustible and non-combustible gases, particulates and oilsoluble compounds. As of recently, the electrical industry is considering measures to reduce substation maintenance costs by moving from time-based maintenance to predictive, condition-based maintenance. This approach has resulted in research and the development of new diagnostic techniques and maintenance tools. The different LTC designs, past and present, pose a non-trivial challenge to the condition assessment of transformer sub-equipment. This problem, facing the electrical and utility industry, can be addressed in different ways, for example by: Increased standardization of new sub-equipment On-board diagnostics, especially for new load tap-changer Moving to solid-state switchgear. Significant progress might be achieved, if the manufacturers could be persuaded to supply each on-load tap changer model with a listing as to type and amount of gas evolved per tap-change, under normal conditions. Background Last year, during the "Ninth Conference", Powertech Labs Inc. provided an overview of the investigations into load tap-changers (LTCs): laboratory findings, results from a field case study and a survey of about one hundred LTCs of different makes and models 1. The results, from thermal runs (Figures 1) and arcing tests (Figure 2, Table 1) provided a characteristic sequence of the oil decomposition compounds, which show commonality as well as significant differences. The analytical work was geared towards the determination of typical gas-in-oil (DGA) compounds, and the development of novel techniques for the determination of unsaturated volatiles (alkenes, alkynes), aromatic hydrocarbons (BTEX) and poly-aromatic-hydrocarbons (PAHs). These sets of compounds were determined by gas chromatography/mass-spectrometry

(GC/MS). The amounts and compositions of particulate matter were determined by inductively coupled plasma spectrometry (ICP).

Figure 8 Schematic for the Sequence of Hydrocarbon Formation from a Mineral-based Transformer Fluid while flowing across a heated "Elkonite"* Surface *Copper infused sintered Tungsten

It was readily apparent that the PAH compounds had no great analytical interest, except in aiding to explain the formation of sludge and coke in LTCs. The study showed that the gas composition (%), due to arcing, is remarkably constant under conditions ranging from those of minute sparks (ASTM D1816), small-scale laboratory tests (needle to sphere breakdowns), as well as load tap-changer and major sustained arcing in power transformer. Significant similarity was also found between the chemical compounds formed during the simulation of thermal stress in the laboratory and our fieldwork on load tap-changers (LTCs). The compounds listed in the schematic of Figure 1 and Table 1 were also found to be present in oil and gas samples from the case study (Kent Substation) as well as in those of the LTC survey. The "gas-in-oil" and the new data of the nine months case study at Kent Substation, describing an LTC under "normal" operating conditions, were remarkably similar to the "normal" operation of most on-load tap-changers 1,2.

Table 1 Formation of Hydrocarbons during Arcing Experiments [Concentrations in ppm (wt/wt)] Arc Time (minutes)

Compounds Propane Propene Propyne Butane 1-Butene 1,2 Propadiene 1,3 Butadiene 2-Methyl Propene 1,3 Butadiyne 1-Buten-3-yne 1-Pentene 2-Methyl Butene Butyne 1,3 Cyclopentadiene Pentane 1-Hexene 2-Methyl Pentene Methyl Cyclohexane Heptane Benzene Toluene Ethyl Benzene M&P Xylene

5

10

15

20

30

60

93

153

213

263

0 0.2 0.1 0 0.03 0.29 0.08 0.04 8.3 0.84 0 0 0 0.23

0 0.36 0.21 0 0.06 0.52 0.16 0.08 16.2 1.76 0 0 0.03 0.45

0 0.63 0.34 0 0.10 0.90 0.26 0.12 22.6 2.76 0 0 0.04 0.69

0 0.78 0.48 0 0.13 1.12 0.36 0.15 26.4 3.71 0 0 0.06 0.94

0.02 1.41 0.76 0 0.23 2.03 0.64 0.23 34.3 6.37 0.09 0.05 0.11 1.70

0.03 2.44 1.39 0.03 0.47 3.55 1.29 0.52 142 12.1 0.19 0.27 0.23 3.40

0.06 4.27 2.37 0.05 0.84 6.07 2.16 0.78 202 22.0 0.29 0.45 0.39 5.20

0.09 6.47 3.66 0.08 1.41 9.3 3.52 1.24 251 36.0 0.50 .77 0.66 7.89

0.14 9.25 5.25 0.10 2.13 12.4 5.11 2.11 306 49.0 0.79 1.14 0.98 11.0

0.19 11.7 6.94 0.17 2.97 14.8 6.96 2.88 338 63.0 1.14 1.61 1.36 14.2

0.17 0 0 3.17

0.18 0 0 2.98

0.18 0 0 3.32

0.18 0 0 3.12

0.20 0 0 3.34

0.19 0 0 3.12

0.21 0 0 3.02

0.23 0.15 0.20 3.07

0.23 0.31 0.32 3.04

0.28 0.38 0.48 3.12

0.66 1.07 2.94 1.16 2.83

0.58 2.20 3.27 1.12 2.50

0.53 3.33 3.64 1.28 0.97

0.81 4.53 3.90 1.24 3.06

0.74 7.62 4.68 1.32 2.96

0.71 15.4 6.24 1.04 3.47

0.67 24.7 8.44 1.65 3.60

0.80 52.0 12.3 2.28 4.18

0.79 66.0 17.7 2.63 4.66

0.71 88.0 21.9 2.79 4.76

Figure 9 Glass Vessel for Arcing Experiments

Transformer Oil Ageing In Oil-Filled Switchgear The ageing of transformer fluids in oil-filled switchgear, such as in load tap changers, is significantly different in several points from the related process in transformers. The ageing of dielectric fluid in transformers is a mature subject and has been described elsewhere in detail 3. This form of ageing is mainly due to the slow breakdown of the fluid accompanied by oxidation, processes which will be accelerated by temperature, the presence of oxygen and the catalytic action of certain materials such as copper. The normal operation of most load tap-changers, in contrast to the operation of transformers, results in the formation of high concentrations of the fault gases. In addition, most LTCs are characterized by the fast accumulation of carbon-rich particulates. The carbonaceous material accumulating in oil-filled switchgear causes maintenance problems and can be removed by filtration. The real difference between ageing in transformers and load tap-changers is a process, which we would like to call "reverse-refining". This process takes place in most LTCs under various operating conditions and will result in the formation of chemical compounds as listed in Figure 1 and Table 1. Among these, the unsaturated and the poly-aromatic hydrocarbons are highly undesirable and problematic. The removal of these types of compounds is one of the main objectives in commercial oil refining process, and that is why we have introduced the term "reverse refining" for the undesired reversal of the process. Reverse refining will lead to an overall lowering of the atomic ratio of hydrogen to carbon∗ and will result specifically in the formation of alkenes (olefines), alkynes and PAHs. These hydrocarbons are highly unstable, and some may polymerize and oxidize spontaneously. A rare example for such processes occurring in transformers is, perhaps, the formation of x-wax. Oil Quality For LTCs (Topic I) Utilities show a great interest in the topic of "Oil Quality" for on-load tap changer. This interest is a result of the realization that LTCs ought to receive more attention than they have in the past. It was and still is, to an extent, quite common to think of transformers when considering oil quality, and only to give minor considerations to the requirements of LTCs. Without filtration, most on-load tap changers are contaminated in a very short time by black particulate matter. This particulate matter adheres readily to critical surfaces within the switching device. Electrical fields and temperature differentials aid this phenomenon. It shouldn’t be a surprise then that the wisdom of filling LTC with transformer fluids of inferior quality is now being questioned. The use of old and recycled transformer oils for service in LTC is quite common and the questioning of what constitutes an acceptable oil quality and good maintenance practice is justified.

Samples (Topic I) ∗

(Examples of such H/C ratios: methane = 4, decane = 2.2, ethylene= 2, acetylene = 1, benzene = 1, anthracene = 0.7, anthracite coal = 0.3 – 0.4. The sludge of load tap-changer is typically around 0.7 to 1.1)

The study included: (1) (2) (3) (4) (5) (6)

Voltesso-35 insulating oil (virgin, naphthenic) Reclaimed transformer oil (essentially Voltesso-35) Insulating oil from an old load tap-changer, which had failed Insulating oil from an on-load tap changer with filtration (Velcon TP-2); the oil was more than four years old Insulating oil from a sister unit to (4) without filtration but similar history, load and number of tap operations A highly refined and hydro treated insulating oil (virgin, paraffinic)

Analyses were performed on: The samples as received (A) After thermal stressing (B) After arc stressing (C) And after thermal stressing followed by arc stressing (D) Only particulate-free samples were used for the 24 hrs-225°C coking test. To answer some of the questions Powertech Labs Inc. designed a novel piece of equipment, a micro-coker. The fluids to be tested were circulated for 24 hours across copper coupons, which were maintained at a temperature of 225±0.2°C. The apparatus is shown in Figures 3 and 4. Drierite Filled Breather

Stainless Steel Lines

Mechanical Press Reservoir 1 Liter (Voltesso, Dried to Specifications)

O - Ring Incoloy Cartridge Heater : 200W Test Coupon

Timer & Clock

Thermocouple

Micro-pump (Speed Adjustable)

Temperature ± 0.2° and Heater Control

Figure 10 Schematic of Micro-Coker The experimental results are reported in Table 2 and Table 3 and supported by the color photographs (handout and overheads). Our investigations revealed that crucial chemical

compounds, which will lead to the formation of surface deposits, are produced during arcing and precipitate on copper contacts at relatively low temperatures (225°C); thermal ageing in contrast will not duplicate this experimental outcome.

Figure 11 View of Micro-Coker with attached 4-point ($ $) Test Leads The elemental compositions and the H/C (atomic) ratios for two of the six LTC fluids and their deposits are listed in Table 3. The compositions are typical of mineral-based middle distillates. The deposits for these fluids (handout) have a lower H/C ratio and are enriched in nitrogen, sulfur and oxygen. Typically, transformer oil sludge has an atomic H/C ratio of ~1. The enrichment of nitrogen and sulfur in the deposit relative to that of the oil shows that organic ‘hetero compounds’ as well as oxidation are important in deposit formation. In the study of mechanisms responsible for middle distillate fuel degradation five processes were found to be involved 4: 1. 2. 3. 4. 5.

Oxidation/auto-oxidation Acid/base reactions Chemical condensation Polymerization: Free radical/ionic addition and condensation Hydrogen bonding

Historically, the first mechanism to be elucidated, oxidation/auto-oxidation, involves the addition of oxygen to diolefins 5. The results from our study have shown that both diolefins and alkynes are in ample supply after arcing (Table 3). The oxygen required for this mechanism to occur is usually present in LTCs. The second mechanism acid/base reactions, is aided is by oxidation. There is sufficient evidence to assume that chemical condensation, the third mechanism, occurs since water and other low- molecular side products are formed in LTCs.

Table 2* Data On-Load Tap Changer (LTC) - Oil Quality, Prior To Coking Resistance Measurements

Test Method → Oil Sample ↓

Acid No.

Additive

ASTM D664 [mg KOH/g]

ASTM D2668 [%wt]

Dielectric Break Down ASTM D1816 [kV]

Color

Coke Res.

Coke Res.

Coke Res.

ASTM D1500 [Intensity]

PLI [mOhm]

PLI [micron]

PLI [Appearance]

<2 4 600 400 <2 <2 500 500 <2 <2 500 Deposit fell off <2 <2 600 400 <2 <2 700 500 <2 <2 500 500

Tarnish Resin Blk. Deposit Blk. Deposit Tarnish Tarnish Blk. Deposit Blk. Deposit Resin Tarnish Blk. Deposit Blk. Deposit Tarnish Tarnish Blk. Deposit Blk. Deposit Resin Resin Blk. Deposit Blk. Deposit Resin Tarnish Blk. Deposit Blk. Deposit

[1] New V-35 [A] [1, A] <0.01 0.12 64 0 2.4 [1] Thermal [B] [1, B] 0.13 0.10 31 6.5 2.2 [1] Arced [C] [1, C] 0.01 0.07 40 2.5 >60000 [1]Therm. and Arc . [D] [1, D] 0.18 0.04 45 5 >60000 [2] Reclaimed V-35 [1] [2, A] <0.01 0.40 31 1 2.3 [2] Thermal [B] [2, B] 0.13 0.18 30 8 3.2 [2] Arced [C] [2, C] 0.01 0.37 36 4 >60000 [2] Therm. and Arc. [D] [2, D] 0.17 0.19 51 8 >60000 [3] Used LTC V-35 [A] [3, A] <0.01 0.20 17 >8 3500 [3] Thermal [B] [3, B] 0.16 0.10 32 >8 2 [3] Arced [C] [3, C] 0.04 0.16 19 3 >60000 [3] Therm. and Arc. [D] [3, D] 0.23 0.04 32 6.5 >60000 [4] LTC V-35 (WFil) [A] [4, A] <0.01 0.31 29 1 3.3 [4 ]Thermal [B] [4, B] 0.13 0.16 26 8 1.7 [4] Arced [C] [4, C] 0.03 0.21 22 4.5 >60000 [4] Therm. and Arc. [D] [4, D] 0.17 0.12 44 8 >60000 [5] LTC V-35 (WOFil) [A] [5, A] <0.01 0.23 28 1 1.8 [5] Thermal [B] [5, B] 0.14 0.12 32 8 1.5 [5] [Arced [C] [5, C] 0.01 0.14 42 3.5 >60000 [5] Therm. and Arc. [D] [5, D] 0.17 0.05 49 7 >60000 [6] Paraffinic LTC TF [A] [6, A] <0.01 0.27 62 0 1.9 [6] Thermal [B] [6, B] 0.01 0.20 38 1.5 1.6 [6] Arced [C] [6, C] 0.01 0.20 45 2 >60000 [6] Therm. and Arc. [D] [6, D] 0.02 0.11 45 3 >60000 Thermal Conditions: 1 Liter TR fluid: 150°C/11 days; Arcing Conditions: 1Liter TR fluid: 5mamps at 7.6 kV/6 hours

Coke Res. = Coke Residue, Blk. Deposit = Black Deposit, Diel.Break, Therm. and Arc. = Thermally aged followed by Arcing, (WFil) = With Filtration, (WOFil) = Without Filtration, TF = Transformer Fluid; Please see also the sample description in the text.

Table 3** Data: On Load Tap Changer (LTC) - Oil Quality, Prior To Coking Resistance Measurements Test Method → Oil Sample ↓ [1] New V-35 [A] [1] Thermal [B] [1] Arced [C] [1]Therm. and Arc . [D] [2] Reclaimed V-35 [1] [2] Thermal [B] [2] Arced [C] [2] Therm. and Arc. [D] [3] Used LTC V-35 [A] [3] Thermal [B] [3] Arced [C] [3] Therm. and Arc. [D] [4] LTC V-35 (WFil) [A] [4 ]Thermal [B] [4] Arced [C] [4] Therm. and Arc. [D] [5] LTC V-35 (WOFil) [A] [5] Thermal [B] [5] [Arced [C] [5] Therm. and Arc. [D] [6] Paraffinic LTC TF [A] [6] Thermal [B] [6] Arced [C] [6] Therm. and Arc. [D]

[1, A] [1, B] [1, C] [1, D] [2, A] [2, B] [2, C] [2, D] [3, A] [3, B] [3, C] [3, D] [4, A] [4, B] [4, C] [4, D] [5, A] [5, B] [5, C] [5, D] [6, A] [6, B] [6, C] [6, D]

Relative Free Radical Activity

IFT ASTM D971 [dynes/cm]

Oxid. Stab ASTM D2112 [minutes]

Unsaturated Volatiles Alkenes [ppm]

Unsaturated Volatiles Alkynes [ppm]

BTEX [ppm]

0.13 1.05 0.12 0.72 0.38 4.17 0.4 2.56 0.29 0.9 0.29 0.32 0.36 1.41 0.43 2.27 1.59 1.72 0.33 1.02 0.05 0.1 0.16 0.08

39.9 18.9 32.8 21.1 43 22 35.1 22.8 30 21 29.5 21.4 38 22 31.5 22.6 38 21 31.7 22.2 44 32 38.3 29.3

215 35 205 70 310 85 280 85 160 30 155 30 255 65 220 65 210 45 200 70 560 250 500 225

<0.1 <0.1 5 11 <0.1 <0.1 8 9 76 <0.1 57 11 <0.1 0.2 19 9 0.8 0.15 17 13 <0.1 <0.1 7 8

<0.1 <0.1 107 223 <0.1 <0.1 177 185 94 0.1 140 188 1.4 <0.1 251 229 2.2 <0.1 254 237 <0.1 <0.1 208 207

<1 <1 67 85 11 <1 89 77 76 <1 158 73 19 <1 160 83 22 <1 134 92 <1 <1 169 80

Oil = [A]; Deposit = [C] C N N O S H/C C H N O S Elemental Composit % [1A] & [1C] 86.93 12.81 <0.3 <0.3 <0.3 72.35 4.46 1.63 22.4 0.52 1.77 ∼18 Elemental Composit % [6A] & [6C] 86.06 14.37 <0.3 <0.3 <0.3 74.95 6.67 0.09 0.35 2.00 **Please see also the footnotes for Table 2, Relative Free Radical Activity is an ‘In-House’ test, BTEX = Benzene, toluene, ethyl benzene, xylenes

H/C 0.74 1.0

Concerning the fourth degradation mechanism: free radicals are more active under elevated thermal conditions (Table 3, column 3) and promote polymerization. A case-in-point is "sample 3" in Table 2 and Table 3. It is the only field sample which shows a sufficiently high concentration of unsaturated compounds (Table 3, data columns 4, 5 and 6) for the mechanism to predominate. The un-saturates disappear completely during thermal ageing (sample 3B, Table 3), but also show up as a resin of considerable four-point resistance (Table 2, sample 3A, data columns 5 and 7). The fifth mechanism helping the formation of particulates is the weak hydrogen bonding. It is safe to assume that a very similar set of mechanisms as shown – for middle distillate fuels –will apply to the chemically related mineral-based transformer oils after they were exposed to arcing and thermal stress in load tap-changer. These initially, mostly loose deposits, if given a chance and high enough temperatures, will decrease in H/C-ratio and become graphite like and hard. The tables, handouts and the overheads demonstrate clearly the effects of arcing and thermal stress. The formation of a poly-aromatic hydrocarbon, "acenaphthalene" was demonstrated under conditions of arcing. This PAH was not present in either the original hydro-treated, paraffinic oil or in the thermally stressed oil but was found in the arced samples. It is a PAH of the formula C12H8 or (C10H6)(CH)2 , it is interesting in that it is composed of two fused rings (naphthalene) joined 1,8 by an olefinic -(CH)2 – bridge ("acenaphthene": C10 H6 (CH2)2 is the compound with single bonds in the "bridge" position). The environment of load tap-changers promotes chemical condensation and polymerization reactions, since most normally operating LTCs will produce the necessary precursors. Particulates, resins, sludge, coke-like deposits are formed and possibly some pure carbon as well. Conclusion (Topic I) Arcing does not affect greatly the oxidation stability of any of the tested oils, whereas the thermal treatment does. In all cases arcing produces aromatic, unsaturated and poly-aromatic compounds, with double and triple bonds. These compounds promote the resin and coke formation in the micro coker during 24 hours of re-circulation over copper surfaces at 225°C. Resin and coke formation can also be observed in load tap-changers. From a thermal ageing point of view the highly hydrotreated, paraffin-based transformer oil appears to be a superior insulating fluid for transformers, but there appears no apparent incentive for its use in LTCs. But, and this is important, for the use in load tap changer, all oils show a severe tendency to coke if exposed to conditions of arcing or a combination of thermal and arcing stress. Under thermal stress, the reclaimed oil was found to be as sensitive to the formation of high ASTM color as are any of the oils taken from load tap changers after more than four years of field service. Thermally stressed, reclaimed oil shows the highest free-radical scavenging

activity among the six transformer oils tested. However, it may exhibit a tendency towards coke formation, which might require more frequent oil replacement than new oil. Old oils need to be replaced after a certain length of service as shown by sample 3A, Table 2, data columns 5, 7 and sample 3A in Table 3, data columns 4,5, and 6 to avoid sudden coking without initial mechanical failure. The "reverse refined" oil might go thermal due to sludging and a reduced heat transfer between the oil-immersed switching assemblies and the surrounding oil. To test for the degree of "reverse refining" one can simply determine the "bromine number" (ASTM D1159). Samples 4 and 5 demonstrate the value of in-line filtration. The samples from the load tapchanger without in-line filtration appear to form more deposit (cf. samples 5C and 5D, Table 2, data column 6). The correctness of this observation for LTCs has to be validated in field trials. The products of arcing, under normal operating conditions, are responsible for the formation of resins and surface sludge. Higher temperatures and the presence of catalyst (copper) accelerate these critical reactions. Development Of High Resistance In Stationary Reversing Switch Assemblies (Topic II) In normal conditions, the reversing LTC switches are moved infrequently. However, a large number of reversing switches on LTCs fail. This experience let maintenance personnel at many utilities to a field practice, which consists of moving reversing switch assemblies deliberately to avoid the occurrence of failure. At the same time, some utilities and equipment suppliers claim that certain contact materials are immune to this type of fault behavior. Powertech Labs Inc. used a modified "North American Transformer/FPE" on-load tapchanger (courtesy of B.C. Hydro) to study coking or the initiation of coking on ‘resting’ reversing switch assemblies. The measuring circuit is presented in Figure 5. The EPRI project specified long-term testing under current loadings between 100 and 900 A, as a high voltage of about 12 kV is applied. The reversing switch contact assemblies were rated for 800A. Since an evaluation of this nature allows for the testing of only three reversing switch assemblies at any given time (three phases), agreement was reached on testing the following assemblies:



Original copper-switch assembly, which came with the LTC New copper-switch assembly (OEM) New silver-switch assembly (HVS)

[*Consisting of CONTACT, REV, SW.R.H. CONTACT, REV, SW, L.H. and CONTACT ASSEMBLY MOVABLE, all: FPE; the switch assemblies were courtesy of High Voltage Supply, Dallas/Texas]

EPRI Evaluation of Contact Materials WO 4204 phase 2

Fluke Hydra Datalogger Model 2625A/WL

120 VAC

12 kV

CT

120 volts

12 kV

Current Injection Adjustment 240 Volts

RGP/Jan00

Figure 12 Schematic of the Measuring Circuit Several parameters for each of the three reversing-switch assemblies were measured at 10 minute intervals, including: Voltage-drop, Current (Amps), Temperature of the stationary contacts, the movable contacts, the oil, and the ambient air. Test runs of one month duration each were conducted at 400 A, 600 A, 800 A and 900 A. Experimentation with the current and exposure time suggested a schedule, which would induce surface and metallurgical changes in the contacts of the resting reversing switch assemblies. This program was implemented and run under computer control. Finally, the oil was drained and the reversing switch assemblies were opened and closed five times. After closure of the LTC and refilling with oil, measurements were taken at current settings of 100, 200, 400, 600, 800 and 900 A during runs of 24 hours duration each.

Preliminary Results (Topic II) Initially, for the experimental range of current and associated contact temperature, the voltage drop increased in this order: new silver contacts < new copper contacts < old copper contacts. The voltage drop for new silver and new copper contacts increased linearly with the current, while for the old copper contacts this dependency was non-linear and of a low confidence fit. After nine months of testing the stationary reversing switch assemblies were opened and closed six times. The order and the linearity remained the same, but the voltage drop for the two copper contacts had increased at the end of the test period and after movement of the same. The silver assemblies maintained the same voltage drop throughout the nine-month test period. The increased voltage drop of the copper assemblies was interpreted as a permanent change affecting the contact surfaces of the copper assemblies: perhaps the formation of deposits was followed by carbonization and interaction between graphitic materials and metal 6. The contact physics and chemistry of silver contacts appears to differ advantageously from those made of copper.

Figure 13 Examples of Current Settings and Voltage Drops

Figure 14 Voltage Drops as a Function of Time Data evaluation is still in progress, but the preliminary indications are as follows: Preliminary Conclusions (Topic II) Silver contacts, aside from the better current capacity, do not suffer the same detrimental "oilcontact surface" interaction as the traditional, lower cost copper contacts. For whatever reason, a "burning-in" (Figure 7) as observed with the new copper contacts, does not occur with the silver-coated surfaces. In the case of copper, capillary forces appear to be responsible to draw the oil in-between the solid conductor surfaces, followed by a burn-in period at higher loadings. We postulate that this is accompanied by high inter-contact temperatures of the insulating oil. The result is the dehydrogenation and carbonization of the oil. This process results in a decrease of the atomic hydrogen to carbon ratio and the formation of a very thin surface deposit. This surface deposit will eventually become graphite-like (for wood-based charcoal, for example, H/C ~ 0.6). Graphite resistance decreases with increasing temperature. As a result (Figure 7) burning-in is observed after a current shift to higher loading. The formation of graphite may lead to the migration of carbon into the copper 6 and a resulting increase of the residual resistance for copper. Indeed, at the end of the nine months

study the ohmic resistance of the new copper contacts has increased. No such phenomena were observed in the case of silver. If these preliminary conclusions are supported by the future findings of our study, this will have far reaching implications, which would apply to metal-to-metal connections in all the oil-filled equipment, including transformers. For oil-immersed connections of any type, access of oil to the immediate contact areas has to be avoided. This would suggest sealing the stationary metal-to-metal contacts prior to oil immersion. Tight clamping would not eliminate voids and hairline gaps between solid copper surfaces into which oil would be drawn even more strongly by capillary forces. Contact coking has been observed in older transformers where load-carrying cables are clamped to posts in the oil-immersed assembly. Acknowledgements This work would not have been possible without the support of BC Hydro Engineering: Messrs Mike Lau, Walter Horn and Joe Holzmann. References 1.

Hans-Ulf Schellhase, Powertech Labs Inc., David Pugh, Consultant, 4580 West 12th Avenue, Vancouver, B.C. Canada, V6R 2N8, "Maintenance-Free Load Tap Changer (LTC)", EPRI Substation Equipment Diagnostic Conference IX, New Orleans, L., 2001

2.

M. Lau, W. Horn – BC Hydro; B. Ward- Electrical Power Research Institute: "Utility Sleuths Detect LTC Contact Wear", Transmission & Distribution World, June 1999

3.

ASTM, American Society for Testing and Materials, R. Bartnikas, editor; "Electrical Insulating Liquids", Engineering Dielectrics, Volume III, pp 341, 1994

4.

Schellhase, H.U., Storage Stability of Middle Distillate Fuels, BC Hydro and Power Authority, Canadian Electrical Association, CEA No. 341 G 469, 1988

5.

Taylor, W.F., Frankenfeld, J.W., "Chemistry and Mechanism of Distillate Fuel Stability", Exxon Research and Engineering Company, 2nd International Conference on Long-Term Storage Stabilities of Liquid Fuels, San Antonio, Texas, 1986

6.

Savio, Leo J., Transformer Consultant; Ward, Barry, EPRI; Crutcher, Russ, Microlabs Northwest; Hanson, Dave, Haupert, Ted, TJ/H2b Analytical Services, "Analysis of Coke Formation on LTC Contacts using Optical Microscopy", Proceedings of the 2000 International Conference of Doble Clients – Sec 8-11, 2000

EVALUATION OF PD ACOUSTIC DETECTION FOR INSTRUMENT TRANSFORMERS INSULATION ASSESSMENT Marceli Kazmierski Institute of Power Engineering Transformer Division •ódz, Poland

Ryszard Sobocki Polish Power Grid Co. Warsaw, Poland

Abstract Usage of the Acoustic Emission (AE) detection from Partial Discharges in oil − immersed Instrument Transformers (IT’s ) for assessment their insulation system is presented. By means laboratory investigations on some models and real IT’s, quantities representing AE, such as frequency, counts, duration, hits, amplitude, were estimated from point of view of their correlation with the apparent charge (in pC) measured by electric method. The acoustic method has been implemented for operated current transformers rated 110, 220 and 400 kV. The measurement may be carried out on live instrument transformers without their switching − off. Acoustic sensors are simply fixed to the earthed steel body of IT’s by magnetic holders. Some of the CT’s with elevated PD level were tested in detail in laboratory and also internally inspected. 1. Introduction Instrument Transformers are one of the largest group of the electrical equipment in a power system. Importance of proper on − line evaluation of their insulation system condition to avoid a catastrophic failure with all its consequences − technical and economic, is out of discussion. PD acoustic detection seems to be useful tool to achieve this goal [1]. In general, signals of the acoustic emission (AE) are characterised by frequency, counts, duration, amplitude and relative energy − Fig. 2. The first problem is to find their correlation with commonly used apparent charge (in pC), measured by the electric method. The paper presents an approach based on a comparison results obtained by means of both discussed methods − electric and acoustic, applied to some simple models and a number of IT’s. Also AE relationship with results of other electric tests and results of oil analysis, that reflect the condition of the paper − oil insulation system of IT’s, is discussed. Moreover implementation of the acoustic method for operated current transformers for 110, 220 and 400 kV is illustrated.

2. Measuring systems Both measuring systems are computer aided and equipped with advanced specialised software for acquisition of the signals, their processing and for presentation of results. 2.1. Electric method Electric system, used for PD measurement in a laboratory, invented and developed in Institute of Power Engineering, Transformer Division [2, 3, 4], is presented in Fig. 1. It consists of two main branches − reference and measurement, separated by inductance Ls.

Measuring block, supplied by DC from battery and located under working voltage, is connected to the registration block by a fibre − optic link.

Fig. 1. Schematic diagram of PD measuring system by means of the electric method. Tt − test transformer, REF − reference branch, MEASUR − measurement branch, Ls − separation inductance, Cr, Cp − coupling capacitors, Zr, Zp − detection impedance, Cx − tested object, MB − measurement block, DC − F − supply unit, S − H − module of peak value detector of „Sample − Hold” type, A/D − analog − digital converter, RB − registration block, RM − registration module, PC − personal computer, CAL − calibration block, Gs − pulse generator, Cs − calibration block capacitor, Opt − fibre − optic link. The system, supported by computer − aided disturbance suppression, enables measuring PD signals below the noise level. The idea of disturbance elimination is based on the comparison of impulses measured in the same time interval in both branches. Elimination criteria are arranged in three steps − amplitude, phase and background noise suppression. Also a self − check procedure is provided by means of supplying the system with a rectangular voltage impulse. 2.2. Acoustic method

For measuring partial discharges by the acoustic method (on a field as well as in a laboratory) AE detection and location system DiSP − 80 of Physical Acoustic Corp. is used. This is an integrated, advanced computerised system consisting of three boards, each having four digital channels. It means it is able to accommodate up to 12 sensors. R 151 type piezoelectric sensors of 70 − 200 kHz frequency band, equipped with special magnetic holders were used. Detail specification of the system and of the sensors is given in [7].

AE signals coming from tested object are compared with adjustable threshold voltage. For their estimation usually five parameters are used − Fig. 2 − counts (of threshold − crossing pulses), amplitude, duration, rise time, and the strength of the absolute value of a detected AE signal, also known as “relative energy” or “MARSE”.

Fig. 2. Measured parameters of AE signals (an idea of the fig. is taken from [5]).

3. Model investigation PD measurements, carried out on real paper − oil insulation systems, show that obtained results are of random character with comparatively low repeatability. Thus it was decided to start the project with investigation some simple models characterised by high repeatability of generated charges. For the comparison tests three arrangements of electrodes were used: flat plates separated by a pressboard sheet, needle − sphere system, acc. to IEC 296 Std., and needle − flat plate. The models were located

inside a steel tank (approx. dimensions − 1100x 600x 400 mm) and oil − immersed. AE sensors were fixed to the tank walls by means of the magnetic holders. An exemplary result, obtained from the needle − flat plate arrangement, is given in Fig. 3.

Fig. 3. Relations between apparent charge, measured by electric method, and AE parameters for needle − flat plate, oil − immersed model. 4. Measurement on substations

Measurements were carried out on live instrument transformers working on substations without their switching − off and the sensors were fixed to the earthed steel body of transformers. Because measurements on substations were focussed rather on detection of PD than on their location, typically four sensors were assigned to one unit. Hence, simultaneously all three phase transformers could be equipped with the sensors that simplified the job significantly. Location of the sensors on CT body for 110 kV is shown in Fig. 4. More than 80 IT’s have been tested on various substations. Generally, they are hair − pin design current transformers rated 110, 220, 400 kV and manufactured between 1967 − 1979. Fig. 5 illustrates exemplary results of field measurements carried out at the working voltage for 110 kV CT, serial No. 409051.

Fig. 4. Location of the sensors fixed to the earthed steel body of a current transformer for 110 kV. Maximum values of some AE parameters for tested population of CT’s are given in Fig. 6. Among them 15 units were selected for transportation to the laboratory of Institute of Power Engineering, Transformer Division for further tests. Unfortunately there was no possibility to take off from visited substations all CT's that showed higher level of AE compared with the level of remaining units. 5. Laboratory measurement

PD's were measured by both the acoustic and electric methods. For their location eight sensors were used. Some results of measurements carried out on the same 110 kV CT, serial No. 409051 are presented in Figs. 7 and 8.

Fig. 5. Results of AE measurement for 110 kV CT (serial No. 409051) carried out at working voltage over field tests. Unstable evolution of PD can be seen in Fig. 7 b). The partial discharges burst and go out. The process repeats several times with varied intensity. However, a quasi − stable level is of order of 300 pC. Position of partial discharges on the sine curve of the applied voltage (Fig. 8) suggests (discharges concentrated on growing segments of the sine curve, i.e. they occurred before its maximum value) that PD sources are within the oil – paper insulation [6]. AE parameters for considered case are presented in Fig. 7 c) − g). Well enough relation with apparent charge (also with its bursting and going out process) may be

concluded for AE counts, energy, duration and in some extend for average frequency. Similar conclusion follows from estimation results of other CT’s.

Fig. 6. Maximum values of some AE parameters measured on CT’ population over the field tests. The next important problem is the sensitivity of the acoustic method. The investigation performed till now shows that the lowest intensity the acoustic method is able to detect is of order of 250 – 300 pC. Nevertheless there are registered cases that the acoustic system could not detect PD’s of 500 pC intensity, sometimes even more. One of the most probable reason is location of partial discharges inside CT against outside location of acoustic sensors. When sources are located deeply in

paper insulation then the acoustic waves may be remarkably damped along their way to the sensors.

Fig. 7. Results of measurements for 110 kV CT (serial No. 409051) carried out in the laboratory. a) voltage over the test, b) apparent charge measured by the electric method, c) − g) parameters of AE measured by the acoustic method

Fig. 8. PD measurement over one period of the sine wave taken at 174 th second of the voltage application from Fig. 7. Next question is location of PD sources; over the laboratory investigations by usage of eight sensors fixed to the CT tank. Fig. 9 shows the location of the PD sources for the discussed current transformer, realised by DiSP – 80 system. A lot of partial discharge sources evenly distributed in a space may be observed. Less of them are located close to sensor 3 and 4, i.e. close to outlets of secondary windings.

Fig. 9. Location of PD sources by DiSP – 80 system in 110 kV CT (serial No. 409051). 1 – 8 Ζ – sensors, + – source location.

Generally differences between AE measurements carried out on field and in the laboratory are observed. They are probably originated from aged oil movement during transportation, influencing a surface condition of the insulation system. Besides PD measurements other electric tests were applied to the CT's transported to the laboratory, namely tg δ, capacitance (C) and 60 sec. insulation resistance (Rins). Moreover an oil sample was drown from each CT's to perform DGA. List of all results for three CT's rated 100 kV are presented in Table 1 and 2. Table 1 Results of measurement on field and in the laboratory on three current transformers rated 110 kV, taken off from one field of Radkowice substation. Year of manufacturing 1973

Serial No.

AE measurement On field**)

HV Lab***)

0 55 25.5 24 184

3.5 100 20 14 3500

Energy Counts 409051 Cnts to Pk Hits Duration

0 108 71 71 320

28.5 700 140 76 28000

Energy Counts 409052* Cnts to Pk ) Hits Duration

0 0 0 0 0

0 560 182 148 4800

409050

Energy Counts Cnts to Pk 7 Hits Duration

Apparent charge

tg δ

pC

[%] 8

375

P − S G

S − PG 9 500

P − S G

S − PG

86

10 P − S G S − PG

6.

Rins

[pF]

GΩ

0.1947 326.6

61.5

4.588 1340

.439

0.1905 313.6

78

4.695 1325

.322

0.2

310.4

5.577 1276.4

85

1.4

*)

CT with no elevated AE measured on field, results for the moment that maximum value of hits was registered, ***) results for the moment that maximum value of apparent charge (measured by the electric method) was registered, P − SG − supplied primary winding, secondary grounded, **)

S − PG

− supplied secondary winding, primary grounded.

Presented results of three considered units suggest that their insulation systems are naturally aged and generally they are in a good condition. Key parameters, such as apparent charge tg δ, C, Rins, content of gases in oil, are still within permissible limits [8]. The conclusion was confirmed by an internal inspection of these CT's. There were no evidences of PD activity.

11 Table 2

DGA results for oil samples taken from CT's discussed in Table 1

Gases H2 − hydrogen CH4 − methane C2H6 − ethane C2H4 − ethylene C2H2 − acetylene C3H8 − propane C3H6 − propylene n − C4H10 − n − butane CO − carbon monoxide CO2 − carbon dioxide

110 kV CT serial number 409050 409051 409052 148 144 148 3 2 2 2 2 2 1 1 1 none none none 4 4 3 1 2 1 3 2 1 50 51 53 906 941 867

11.1

Air

Sum of flammable gases 11.2 Total sum of gases

93782

93551

92722

212

208

211

94900

94700

93800

6. Conclusions

Discussed investigation has not been finished yet. Till now the following may be concluded: •

The acoustic method of PD measurement is easy applicable to field tests carried out on live IT's (without their switching − off), and is safety and simple in practice.

•

AE parameters, such as counts, energy, duration, and in some extend average frequency are correlated well enough with the apparent charge (in pC). It seems that they should be interpreted together. This needs further investigation supported by computer processing of results.

•

Sensitivity of the acoustic method is of order of 250 − 300 pC provided that PD sources are not located deep in paper insulation. In that case the sensitivity is worse up to approximately two times.

•

There is no problem to detect and to assess a serious fault in IT insulation system by means of the acoustic method. If the system degradation is low, its evidences are not easy to estimate, among other due to low repeatability of PD phenomena.

7. References

1. Skubis J.: Acoustic emision for testing electric equipment insulation, IPPT PAN, Warszawa 1993 2. 3.

(in Polish). Arcab M., Dalek J., Dymowski S., Jezierski W., Kozlowski M., Wróbel R.: Arrangement for PD measurement, Patent notification No. PL 172289 (G01R 29/24) (in Polish). Kozlowski M., Jezierski W., Dalek J., Dymowski S., Arcab M., Wróbel R.: Method and system for PD measurement, Patent notification No. PL 172269 (G01R 29/24) (in Polish).

4. Kozlowski M.: New method and measuring system of improved PD evaluation, including computer − aided elimination of disturbances, Mat. Ninth International

5. 6. 7. 8.

Symposium on High Voltage Engineering, August 28 − September 1, 1995, Graz, Austria. Pollock A., A.: Accoustic emission inspection, Metal Handbook, Ninth editon, vol. 17, ASM International (1989), pp. 278 − 294 (copy included in [7]). Partial discharges in transformer insulation, by Task Force 15.01.04, CIGRE, Session 2000. DiSP − 80 User's Manual, Rev. 0, June 1999, Physical Accoustic Corp. Maintenance Manual of Power Transformers, Energopomiar − Elektryka, Gliwice 2001 (in Polish).

TRANSFORMER CONDITION ASSESSMENTA NEW APPROACH

Alan Wilson, Pedrag Vujovic Doble Engineering Co, Watertown, Massachusetts, USA John Anderson Xcel Energy, Minneapolis, Minnesota, USA .

ABSTRACT The development by Doble Engineering Company of methodologies for condition assessment for power transformers is discussed. It uses the best information and data available for units in North American utility systems and is focused on risk for ongoing operation. This is an important process likely to receive greater use in the future, particularly for units with high financial exposure in supply contracts. The Doble condition and risk assessment methodology is based on a proprietary “risk scoring” system applied simultaneously on individual units and group of transformers of the same design and similar operating conditions and age. The methodology requires a statistically valid sample of performance data for transformers and main components. Doble’s Equipment Performance Database is used but the need for reliable and early detection of changes in trends and failure mechanisms requires further cooperation with utilities to expand the existing database into a cost effective and industry-wide equipment performance one. Communication and database integration technologies are available and already used within electric power industry. 1. INTRODUCTION Recent reviews have focused on the age and adequacy of the supply system and its equipment [1,2]. For much of the system, installations at 110 kV and higher started in the 1950s and continued into the 1970s; thereafter the rate of new investment began to decline. Not only is the age of a good proportion of the major equipment in use throughout the developed world very similar, it is not far short of the estimates of asset lives assigned at the time of construction [1]. In some companies this lack of recent investment is coupled with power shortages and the companies are locked into tight commercial supply contracts. These issues

have been highlighted in the recent publication of the Cheney report “A National Energy Policy”[2]. Faced with an aging asset base and capacity restrictions the major issue is to understand the risks for continuing life, possibly at higher load levels, of key items of power equipment- and particularly those with high capital cost and long lead times for replacement. Common topics when considering asset management and maintenance strategies for aged power equipment include asset condition, asset health, asset remaining life and life extension, and perversely – overloading of circuits. The underlying issue is an evaluation of risk from further use under various operating conditions. Ideally, it needs to be addressed on two levels: at “system level” and at “unit level”. The “system level” approach involves taking a broad view of company exposure when assessing group of equipment with specific and critical application/function, such as GSU transformers, “sub-transmission” or transformers in “critical” transmission/distribution substations, for example. At “unit level” the methodology involves establishing health and risk for a specific unit. The best results and the highest reliability of results are when both condition and risk assessments are performed at both levels simultaneously and in a coordinated way. The process requires an evaluation of the current status of the technical condition, strategic and financial exposure of a group and of a unit, made within the context of the initial constraints set by the design and specification, past use abuse and maintenance/repair/ refurbishment, assessment of past and current operating condition relative to known generic weaknesses of the design group and failure rates, for example. Several groups have developed methods to evaluate condition, and some earlier Doble Conference papers describe these [3,4]. Within the Doble Company a systematic “scoring system” process is used, considering each element and leading to a weighted risk score for FMMAA (Family/Make/Model/Application/Age) and a weighted and “relative” condition and risk for specific unit itself.

2 2.1

FEATURES OF TRANSFORMER LIFE AND RISK ASSESSMENT The idealized situation

In 1993 CIGRE identified three different criteria determining the end of life - technical, economic, and strategic [5]. Financial or economic life is an accountant’s measure used to quantify asset value in company accounts. Whilst the consequences of failure will have a capital cost of the current asset value, the failure can have other negative effects of a strategic nature. It could have a negative impact on the community; share value could suffer fueled by adverse reporting.

Technical life is when the condition of any integral part is beyond further service and the cost of refurbishment becomes unreasonable. Life is normally associated with age, but this is not always the case. Generally transformers are expected to deteriorate slowly, eventually reaching a condition of the “onset of unreliability” and this could also force an end of life decision. When systems were installed most utilities thought 35 – 40 years was a reasonable expectation of life, particularly for transmission transformers. A key consideration was the life of the winding insulation, where the dominant aging factor is time at operating temperatures. More heavily loaded power station units were often given a shorter life. Experience has shown that for some designs, in some applications, units do follow this pattern and can be expected to last 50 – 60 years or beyond before such old-age failures occur. For many utilities their holding of sufficient units of the same design is too small to extract meaningful statistics. This does not apply in NGC, the transmission company in England and Wales, and their failure statistics were reviewed in a paper at the 1997 Doble Clients conference, [3]. For their population of almost 1000 transmission class units the failure patterns could be segregated by design group families. Most families fell into the classic bathtub pattern of a very low failure rate of 0.3-0.5% per unit per year for most of their life, and Monte Carlo analysis indicated that the rate will increase at ages above 35 years and the average life should be in the region of 50-60 years.

2.2

Reality

In many parts of the world the experience relates to a much higher rate for early life failures, with a random failure rate, an average age at failure in the range 15-20 years, and failure due to modes other than paper aging. This presents a more complex model for technical life than originally considered above. Analysis indicates very few of these families will reach old age and fail due to normal winding aging. Usually the failures can be attributed to some design weakness- short circuit withstand, overheating/ aging due to poor control of leakage flux etc. This failure pattern is consistent with the results of an international survey carried out by CIGRE [6], by papers produced by US insurers [7]. Various US national and international studies report the average life at failure to be around 15-20 years and with a high random failure rate. Within the main tank the key areas of concern relate to mechanical changes caused by short circuits and also various degradation processes associated with moisture levels in the paper, barriers, and oil. Another major cause of failures is in the bushings and tap changers. One way to avoid many of these early life failures is to have an understanding of the performance capability of main tank, LTC and bushings. The areas of weakness can be revealed in surveys such as those carried out each year by Doble.

CIGRE Substation % WINDINGS 19 CORE 3 TANK& OIL 13 LTC 41 BUSHINGS 12 OTHER 12 Period 1968-78 Number of failures 1000

DOBLE All designs % 43 4 8 16 19 10 1993-8 1500

DOBLE One design % 47 7 12 8 8 1980-2000 100

Table 1. Causes of failures, from CIGRE [6] and Doble From this table it can be seen that failures due to LTC faults are much less in the Doble review while windings failures are much higher. This may be due to the greater understanding of LTC issues since the 1983 CIGRE experience or the LTC design but could be also due to change in failure mechanism attributed to the change in age or different/changed operating conditions. In the Doble data the pattern for one specific US design is compared with the total population. Here there are greater problems with the winding, but these were, probably due to “known” electromagnetic design issues. It is clear that asset/life management strategies based on the numbers such as presented in Table 1 could be significantly different depending on the data used. It is also clear that asset and maintenance management decisions in today’s changing and tomorrow’s even more competitive market should be based on statistics that are more “specific”, “up-to-date” and reliable. Reality is that for many utilities their holding of sufficient units of the same design in similar operating conditions is too small to extract meaningful statistics. Reality is also that probability of timely (early) detection of increased rates of non-random failures associated with design weaknesses, changes in failure mechanism caused by age/wear or changes in operating conditions is directly dependent on the total amount of data collected.

2.3

The roles of condition and risk assessment

Condition assessment is an important step in asset management with many aspects from strategic and day to day operation.

•

Condition and risk assessment - ”life assessment” - is used to enable robust capital investment programs, to identify relevant maintenance tasks and their timing - even down to the unit level, identifying maintenance costs and cost-justifying maintenance resources. Operation and maintenance departments are being pressed to justify their role as senior managers seek to establish a link between the level of maintenance and system performance. An important final reason is to reduce blind risks by using life assessment to define the exposure.

•

Life assessment is turning condition assessment to a failure risk assessment, combining technical, financial and system aspects.

•

Life assessment should be an essential tool for identifying opportunities to overload or extend life, and monitoring progress during periods of such operation.

•

Life assessment should be an ongoing process. It is an integral part of asset management, allowing control of lifetime costs, maintenance needs, effectiveness of expenditures, and a means to identify incipient failure processes.

•

Life assessment is more than just testing. The engineer has to look at a whole range of information. This includes data and reliability measures. It includes system requirements - how the unit was used and when it was first bought- how it is currently being used and how it is intended to be used. Diagnostic data is used to look for trending, benchmarking against other units. The final step is a risk assessment to identify the exposure to technical, financial, and strategic risks.

3 DOBLE’S CONDITION ASSESSMENT: THE TWO STEP PROCESS Doble’s Condition Assessment program is a two-step process. Both steps include proprietary risk “scoring“ system and combine analysis of individual units and FMMAA analysis (family /make/ model/ application/ age) of similar designs with similar operating conditions and age. FMMAA analysis is based on Doble’s existing equipment performance database containing test results and equipment failure and trouble data collected from Doble’s customers for over 40 years. 3.1

“Scanning” as a low cost approach or a first step

It would be often too difficult and too time consuming/costly to undertake comprehensive life assessments of every single transformer in a transmission or distribution system. The “scanning” approach is suggested as a low cost assessment and/or step to provide “initial” risk assessment and ranking. This should identify the group of units that are in a sound condition and at a low risk due to their technical condition. The remainder, those identified as higher

risk can then be selected for more detailed investigation, as identified in the following section, 3.2. The Stage 1 is shown in Figure 1. The first step is essentially an office-based review of available information combined with selected in-service tests when justified. These include: •

Basic nameplate information will identify the manufacturers and vintage. From this, the actual designs can be identified, together with details of all accessories, including bushings that have been used. The purchase documents should include design information such as the winding configuration, accessories, materials,manufacturing processes, and standards used performance.

•

Problems found either in manufacture or factory test with these units are identified and assessed against service experience. This might include poorer performance in tests, such as for thermal rating.

•

User circumstances defined in the original specifications for ambient temperatures, load, voltage, power factor, source impedance, lightning levels, short circuit withstand, acceptable losses, source impedance and load power factor may now have evolved into ones where these requirements are inadequate for the current operating environment. The purchase specification is, therefore, an important document to receive and review in light of current manufacturing standards, and against the actual operating environment. Client discussions and documentation should indicate any significant requirement to operate the unit above nameplate ratings, whether system voltages are at the top of the voltage range and over-fluxing is a possibility, and whether fault frequency or levels are higher than specified. The latter could occur if source impedance or ground impedance have changed, or protection is slower than specified.

•

Service problems with the design family reported to the manufacturer, and requiring an advisory, are identified. Reports of failures on similar designs (main tank, bushings, surge arresters and LTCs) are a valuable source of information- being able to relate trends in test data for all units to the failure and rates of aging revealed in the failed unit. Details of any major rebuild are identified, and related reports often give a good indication of the rate of aging generally to be found.

•

A review of the Doble failure and problem database for each of the design families of transformer and accessories will yield the type of problems shown in Table 1 for the design group.

•

In-service diagnostic tests are reviewed for trends, and benchmarked against the population of units of the same design group.

On the basis of such a survey the population can be identified by design groups. In the Doble process this information is scored against a failure mode and effects tabulation. Included in this are the diagnostic parameters that are indicative of each type of malfunction. Identification of key performance problem areas provides the means to identify what diagnostic tests are important for review for indication of each of the malfunction types. For this analysis the original test data for the unit and accessories are used for comparison with the specification and also with in-service results. The company has extensive databases that can be used to identify normal significance of deviations for values found in the test group. Data for individual units will also be reviewed for changes throughout life, and the current rates of change. For historic tests the diagnosis generally focuses on insulation quality and winding and connection integrity. These tests are: • • • • • • •

Winding power factor: this indicates the extent of moisture or contaminants on the winding Winding capacitance: this can indicate gross winding movement Bushing power factor/ capacitance: this can indicate deterioration and risk of failure Insulation resistance: this usually is an indicator of oil contamination Oil tests: these will indicate oil quality indicating needs to refurbish or change the oil. This will avoid the risk of advanced winding aging, sludge formation leading to overheating and reduced electrical performance. Dissolved gases: these give indication of the presence of incipient failure processes. Sometimes oil analysis with Furan levels is reported, and this indicates (usually localized) paper degradation. Winding resistance and turns ratio: these indicate the integrity of the winding and connections.

The aim of this process is to score the key design and service performance areas for overall performance, and evidence from diagnostics. This will yield a score and position in a ranked list for technical condition for each unit. This list can be modified with a criticality factor assigned to each assessment area, turning the condition into a technical risk factor. Low risk units will be selected to receive some in-service non-invasive testing, such as a new oil sample, infrared testing of bushings, infrared scanning of the LTC, and perhaps parts of the main tank. Other such tests may include partial discharge testing using contact probes and using UHV radio interference methods. This range of tests will indicate problems within the bushing and within the main tank. This would be complemented with a comprehensive visual inspection. These are all fairly rapid tests and inspections, and this would be done on a substation basis.

Client Review Identify Assets, Documentation, Scope of Work, Accept Anecdotal Evidence

Client Design Related Documents

Client Unit Level O&M History

Client Unit Level Test History

Client Failure and Defect Reports

Doble DTA/Oil Files for Benchmarking

Doble Failure and Defects Reports

Doble New Oil Test

Doble On-line Diagnostics

Doble Fault and Defect Chart and Review

Identify Design Related Deficiencies

Identify Unit Level Deficiencies

Stage 1 Prioritized Assessment of Units

Figure 1 Doble Transformer Life Assessment Protocol - Stage 1 Technical Assessment

3.2.

Comprehensive analysis for “critical” and high-risk transformers

For the higher risk and “critical” T & D units and generator step-up units, a range of off-line testing would be arranged for stage 2, looking at the dielectric, electrical and electromechanical properties. The review will include a design assessment, looking at how the manufacturer needed to modify design practice over the years, and a full review of a specification originally used on how this has changed. Insulation life is assessed from time at temperature records. For a 30-year-old unit it is impractical to calculate this comprehensively. Data would be reviewed for consistency and

ten or more months selected as representative of lifetime loading. Loss of life is calculated, ideally from hot- spot records. This is then modified using an estimate of paper moisture content. This can be measured on line using systems such as the Doble Domino [8], or estimated from saturation measurements of oil moisture and partition curves.

Stage 1 Prioritized Assessment of Units

Review and Release of Documents with Client

Review Client O&M Policy

Review Client System and Unit ; Performance

Doble Offline Testing

Assess Strategic and Financial Performance

Doble Life of Insulation

Doble Fault and Defect Chart

Technical Condition Review

Life Assessment

Stage 2 - Prioritized Assessment of Units

Outcomes of Stage 2

Figure 2

Doble Transformer Life Assessment Protocol - Stage Two Life Assessment

A full range of diagnostics and inspections would be performed- assessing not only the electrical insulation but also mechanical condition and unit integrity. With this range of data, the higher risk transformers can be scored for risk of failure in a 5-year period. From this, life management recommendations can be produced. Again, a proprietary risk “scoring” system will be used for combined unit and FMMAA analysis using Doble’s equipment performance database.

3.2.1

Electrical tests

Most utilities will only have done the limited number of the tests listed above in section 3.1, and so the assessment will be incomplete. Particularly important are assessments of the mechanical condition of the winding and core. This is achieved by Doble using their sweep frequency, leakage reactance and exciting current test apparatus and methods. The full range of tests is considered to be the following: • • • • • • • • •

Swept Frequency Response Analysis Leakage reactance Excitation current Winding Power factor/Capacitance Bushing Power factor/Capacitance Surge arrester watts loss and counts Turns ratio Winding DC resistance Core ground resistance (for core form designs).

These tests will be used to assess the condition of: • • • 3.2.2

Mechanical condition of core and windings Connection issues in the winding or in core ground Dielectric quality of main tank components and bushings.

Oil tests

A sample would be taken and analyzed in the Doble oil laboratory. DGA, water, particle count, color, inhibitor, Furans, Neutralization Number, Interfacial tension, Dissolved metals, power factor, Dielectric breakdown voltage, viscosity, specific gravity, PCBs and moisture in paper. These tests will be used to indicate: • Oil condition indicating the need for regeneration (filter and drying), replacement, or inhibitor addition • Metals from pump wear that could affect dielectric strength and lead to flashover • Furanic compounds indicative of overheating in winding paper

• •

Incipient failure process- partial discharge, arcing, overheating Calculation of paper moisture content.

. 3.2.3

External inspection

The units are out of service and the usefulness of this inspection is less as a consequence. It may be possible to activate the cooling systems, and this would be done. The external condition of the cooling equipment for rust, leaks and debris can be made. Information on the method for ensuring no water leaks into the oil would be gained. Bushings would be inspected for leaks, and damage to sheds. The main tank would also be checked for leaks and rust. The moisture control system would be checked for effectiveness. Control devices such as gas collector and sudden pressure relief systems would be checked. The control cabinet would be examined for water ingress, heater effectiveness, wiring condition and unusual appearance of breakers, contactors, relays, terminals etc.

3.2.4

Internal Inspection Internal inspection as part of a life assessment is at its most effective when associated with a collection of units having high service years and where one is selected as being the most aged, and is dismantled. Full access to all component parts is then available. A true assessment of cooling performance, rates of conductor and its insulation aging can be made from laboratory samples taken at the optimum locations. The design can be assessed for local overheating, adequacy of flux shields, core condition, clamping arrangement and the extent of loosening. When the examination is made on units to be (possibly) returned to service the inspection is less effective and further damage can be introduced- from debris, ambient moisture and mechanical damage from the inspector gaining access. Much of the internals are hidden by construction of the windings, and by the insulation sheet materials used around the windings. Where an internal inspection has to be made in this situation the focus is on general condition – of the tank, connections, shunts, core joints, winding and clamping. Evidence of free water and particles, particularly in the sump, should be assessable. Evidence of winding distortion may be seen, particularly if the outer winding is affected. Distortion of the LV winding is the most common and that is rarely seen, but may be inferred from displaced support blocks.

Examination of the structures at top and bottom of the windings is consequently one of the most important areas for inspection. The condition of the clamping, loss of tension and winding looseness are important. Contraction of paper during aging can cause loosening, as can in-service short circuit forces. This will limit the capability to withstand further short-circuits. Leads are the most accessible areas of insulated conductor. The outside of the winding can be examined for any movement of the insulation system around the winding, loss of paper, or overheating of paper.

Lead connections are important, particularly with old units, where the brazing can deteriorate leading to a hot spot. Grounding straps can also be examined for similar deterioration. Some inspection of cooling ducts is possible, albeit limited. Better information on deterioration in cooling performance, due to blocked cooling ducts, for example is obtainable from a thermal diagnostic developed recently by Doble [9]. (This system is installed on in-service units and effectively reconstitutes the factory heat run, thereby determining the loss in performance). A paper sample would be taken from a lead in a low stress area for measurement of paper degree of polymerization. The lead would then be repaired. Clearly future conditions will impact the strategic assessment- load levels, voltages, over voltage transients, short circuit levels, and environmental impact. Financial factors will include loss evaluation, maintenance costs, outage incidence and costs, safety, cost of constraints are separate issues for scoring and evaluation.

4

CRITICALITY AND RISK

With each of the foregoing assessments it is possible to come up with a score for the condition. The real issue is whether the aspect under review is critical. Usually this means whether the transformer is likely to fail as a result of a condition worsening. The approach is to moderate the condition score with a weighting factor. The ascribing of these factors will differ between experts, but judged on their opinion and experience. This may vary on application. For example oil leaks at gaskets is a poor condition that is unlikely to lead to immediate failure. But in some situations any measurable oil loss may be environmentally unacceptable, requiring a higher criticality factor. Risk is usually defined by consequence and probability. The approach being introduced into IEEE circuit breaker and transformers committees are decision matrices where events are assigned into a matrix of consequence and probability. Both are difficult parameters to ascribe. Probability relates to statistics, where greater cooperation between utilities is needed. The industry needs to work towards an agreement of terms and consequences (this is being reviewed within CIGRE SC 12 and 23). Foot example “catastrophic” may be described as an event that threatens: • • •

Financial exposure threatens long term survival of the organization and share value- this could easily occur with a generator step-up failing at times of high spot prices and punitive contracts. Fatalities- in 2001 three deaths of workers following a transformer failure was quoted as one reason for the demise of a utility. Federal or state enquiry

•

Major exposure to worldwide news/ TV.

Other descriptions may be developed for “Major”, “Moderate” and “Negligible”. This is clearly an area of future debate as focus changes from purely technical considerations.

CONCLUSIONS Life management of transformers is an ongoing process focused on identifying levels of risk of failure, maintenance requirements, and replacement strategies based on technical, strategic and financial considerations. Condition assessment is an essential part of the process and uses a range of diagnostic indicators to reach assessment of the overall technical condition of the unit. Doble’s approach relates condition to a “criticality” in terms of life expectancy and risk of failure for individual units. This is preferably done in combination with analysis performed for group of transformers of similar design (family/make/model), and operating conditions (application) and age. This process requires industry collaboration in collecting a statistically valid set of transformer performance and operational data. With relatively high average age and increasing loading, the need is now even greater for collaboration between groups of utilities with similar design families to pool data for early identification of changes in patterns indicative of increased unreliability associated with approaching end of life for a specific family/make/model/application/age groups.

REFERENCES

[1] CIGRE Brochure 176, 2000, “Aging of the System, Impact on Planning” [2] Cheyney, R, and Lundquist R, 2001, “National Energy Policy Report, Washington, USA [3] Jarman, P. L., Lapworth, J. A., Wilson, A., (1997) “Life Assessment of 275 and 400 kV Transmission Transformers,” Proceedings of the 1997 International Conference of Doble Clients, Boston, MA (USA). [4] Perkins M et al, “Transformer Life assessment tools and methods” Proceedings of the 2000 International Conference of Doble Clients, Boston, MA (USA). [5] CIGRE (1993) ELECTRA No 150, October issue, pp47-48

[6] CIGRE (1983) ELECTRA No 88, May issue, pp47-48 [7] Bartley, W. H. “An Analysis of Transformer Failures - A 20 Year Trend,” Proceedings of the 2000 International Conference of Doble Clients, Boston, MA (USA).

[8] Lewand, L. R. and Griffin, P. J. “Development and Application of a Continuous Moisture-In-Oil Sensor,” Proceedings of the 2000 International Conference of Doble Clients, Boston, MA (USA). [9] Lachman M.F.et al, “Experience with dynamic loading and thermal analysis of power transformers.” Proceedings of the 2001 International Conference of Doble Clients, Boston, MA (USA).

Forensic Failure Analysis: How Materials Fail in Load Tap Changers E. R. Crutcher III Associate of ADAPT and TJH2b ABSTRACT The failure of a Load Tap Changer (LTC) is the final act in a long series of events that began as an irreversible change in one or more of the materials used to build the LTC. These irreversible changes are the true root cause of the failure. Discovering these material changes in LTC components as part of a failure analysis can save millions of dollars. These savings are the result of avoiding future failures in the same type of LTC and by identifying failure mechanisms that are the result of a design characteristic that may be present in other LTC's. Carbon build-up is not a failure mechanism. The condition that led to the irreversible changes in the components that caused the carbon build-up is. The conditions that cause the changes in the materials in an LTC leave a history written in the surfaces and the internal microstructures of the components. This history can be read using microscopical analytical techniques. This paper documents how fatigue failure, stress relief, and mechanical distortion induced by thermal control problems have led to LTC failures. It also documents how forensic materials sciences can be used to determine the sequence of events leading to the final failure.

INTRODUCTION The goal of failure analysis is to determine the cause of the failure in order to control future liabilities. A LTC fails because the configuration of its material components can no longer support the function for which the system was designed. The failure is in the materials. Everything else is a symptom of the failure. An important part of failure analysis has to be the analysis of the materials that constitute the components involved in the failure. The irreversible change in those materials that lead to the final failure may have occurred long before the final failure. The signature of those changes is generally still present even after the failure. Specific measures can be taken to control premature failures by accurately identifying that signature and the cause for the changes. Materials change in a variety of ways as a function of their environment. Atoms migrate, chemical reactions occur, use imposes stresses that change surfaces and structures, metals slid against one another, arcs occur, heat is generated, cycles are repeated, in brief: things change, mechanically, chemically, and structurally with time and with use. The history of a part is recorded in these changes. Some of these changes are clearly visible to the unaided eye when an LTC fails in service. Determining which changes are a result of the failure, which lead to the failure, and which are not related to the failure can be a difficult task. When these materials are examined under the microscope changes in their microstructure indicate the conditions that existed leading up to the fail. The materials analyst can identify these changes

and the approximate time of their occurrence with respect to the failure. This type of analysis can be expensive if not targeted at specific components of interest. All materials change but the analysis should be focused on specific materials involved in the failure. The materials analyst needs to know as much as possible about the mechanics of the system being examined and the sequence of events that lead up to the failure in order to restrict the scope of the analysis. Proposed scenarios are also useful because they can be tested by looking for the presence or absence of evidence for the events that had to occur with the given scenario. The analyst also needs the failed parts and associated structures, optimally along with similar parts that did not fail. Well-documented photographic records of the failed unit and diagrams are very useful. The more information provided the analyst the more information that can be gathered in the process of analysis. Care should be taken to minimize the handling of the components prior to submitting them for analysis. Some handling is necessary and should be documented so that intentionally induced damage of components can not be misinterpreted as related to or associated with the failure itself. A failure analysis has much in common with the investigation of a crime scene. The event has already occurred. The intent of the investigator is to determine what happened, when (in sequence) it happened, and why it happened. Toward that end the trace evidence at the scene is often critical. The forensic analyst can often provide the information needed if the evidence is provided for analysis. EXAMPLES The examples that follow stress the use of materials information to solve field failure problems. These are not case histories for the most part but rather examples of how materials provided clues as to the nature of the conditions that existed leading to the final failure. 1. HOW THINGS CHANGE CHEMICALLY Some materials change in ways that can be easily documented microscopically. Consider the practice Copper Oxide Free Zone of bonding a silver/copper alloy contact plate to the Near the Surface of the surface of a toughened copper contact. The photograph at right is a cross-section of a copper oxide toughened electrical grade copper contact. The copper oxide toughening can be seen as a Copper Oxide distribution of very small copper oxide grains (black Toughened Zone specks) in the boxed area of the photograph. Near the surfaces of the copper contact there was a zone of no copper oxide. This pattern is typical of the phosphate treatment given to this type of copper alloy to intentionally remove the copper oxide from the surface zone of the copper. The phosphate process leaves a trace of phosphate behind. The photograph below shows a bond region where the copper contact was bonded to a silver/copper alloy using a silver copper brazing

material. The result was a trace of phosphate left at the surface of the copper. Over time the residual phosphate reacted with the copper in the brazing alloy forming a copper subphosphide, sub-oxide phase. The dark color of this material in reflected light relative to the copper and silver testifies to the higher resistance of this material. The result is heating. As this film grew the heating increased until a temperature in excess of 250 degrees Celsius was reached. At this temperature carbon deposits began to grow between the contacts. As the carbon films grew the temperatures climbed and finally resulted in a failure Silver/Copper of the component. The cause of the failure was in the process used to make the contact and in the lack of an adequate heat dissipation design to handle the heat loading created over time. A change in the materials or the heat dissipation design probably would have prevented the failure. If the contact had not been cross-sectioned and analyzed the original source of heat would never have been identified. The Coppe reaction had not occurred when the contact was produced so testing of the new contact would not have indicated a problem. The problem occurred over time. 2. HOW THINGS CHANGE PHYSICALLY Compressed, resin impregnated wood provides an example of how things can change physically over time. This material has been used in the past to construct housings, spacers, and supports in load tap changers and other equipment. While this makes a good insulative construction material it is prone to distortion with time due to the anisotropic distribution of stresses. Wood is structurally different along each of its three axes: radial, longitudinal, and tangential. The photograph at right show a crosssection of a resin impregnated compressed wood LTC part. The radial direction is N-S and the tangential direction is E-W in the photograph. When it is compressed there is more distortion radially (note crushed region) and tangentially (note horizontal dislocations) than longitudinally. The tangential distortion is in shear while the radial distortion is in compression. The result is that the mechanical rebound is greater in the radial direction than in any other direction. Shafts that began with a

circular cross-section become elliptical in cross-section with time. The radial direction in the cross-sectioned shaft in the photograph at below is approximately E-W. This can result in mechanical problems if tolerances are close. Another problem is that parts made from wood are often not perfectly aligned to the grain of the wood. The result is that the part warps. This also can create mechanical problems. The asymmetric rebound in compressed wood should limit its use to dimensionally noncritical applications. Distortion over time in wood is even a problem if the wood has not been compressed and impregnated with resin. Differences in the swelling as a function of moisture and direction in the wood limit the use of wood to applications that are not spatially critical or sensitive to slight distortions. 3. HOW THINGS CHANGE MECHANICALLY Many metals and metal alloys are sensitive to fatigue stress crack failure. In these materials cyclic bending of the metal creates cracks that grow with each cycle of bending. A silver alloy ribbon conductor used in a load tap changer resistor is a good example. This resistor was designed as a flat plate insulator, about three inches wide and nine inches long, around which the ribbon conductor was wound. As the resistor was used the conductor became hotter and expanded, primarily along the length of the ribbon. The expansion resulted in a maximum stress at the center of the long run of the conductor across the insulator. With each cycle cracks began to grow on the outside curvature of the conductor. As the cracks grew the local resistance increased and the cracks grew quicker. Finally the ribbon separated and the system failed. The reason for the failure was uncertain until the fatigue stress cracking was seen in a cross-section of the conductor under the microscope.

4. WHAT HAPPENED FIRST The sequence of events leading to a failure can often be partially interpreted by the position and nature of the debris in the area of the failure. In one tap changer the drive shaft bearing seized and the contacts were caught in a position that resulted in severe arcing and finally in meltdown of the contacts. The issue at hand was what had caused the bearing to seize. An examination of the bearing resulted in two interesting pieces of evidence. First, the outer bearing race had become magnetic. Steel can become magnetic as a result of repeated, heavy, unidirectional loading. This tended to retain wear metals in the race rather than allowing them to be washed free with the lubricating oil. The second was a significant population of flattened copper and silver arc debris particles (see 40X photograph at right). This indicated that the bearing was still turning at a time when significant amounts of arcing debris where being generated. The bearing seized as part of the final event but a misalignment of the

contacts resulting in significant arc was apparently an on going process prior to the bearing failure. The failure mode was already underway before the bearing seized. The seizing of the bearing initiated the final sequence of the failure. Though the bearing failure in this case was the result of a failure process that was already underway the examination indicated that the bearing was undersized for the loading it experience in use. The metal should not have been magnetized. The bearing was operating in a chronically overloaded condition. The bearings are significantly larger in the next generation of this model. 5. WHEN DID IT HAPPEN When did the conditions that led to the final failure first appear? Which conditions where already present, and which were the result of the failure? How long did a condition exist prior to a component's final failure? These questions can often be answered in at least relative terms by an analysis of the failed parts as seen above. One aspect of timing involves the difference between a chronic condition and an intense exposure to heat. A thermal gradient is created in a part that is exposed to heat over a long period. This gradient causes different levels of damage as a function of distance from the heat source and the nature of the material being heated. There are critical temperatures at which some material change in characteristic ways. That could be a melting point or a point at which a typical chemical degradation occurs. This question was relevant in an analysis of a load tap changer that had a polymer composite material forming the housing for the movable contact. The housing was crosssectioned and examined to determine the depth of penetration for the damage and the nature of the damage as a function of depth. The pattern indicated that heating was a chronic condition. The extent of damage, as indicated by increased darkening of the material, was progressive and exhibited a pattern that was consistent with a relatively slow thermal diffusion rate. Superimposed on this pattern was a severely charred area in direct contact with one of the contacts that had melted during the final failure. This damaged area was uniformly charred and cracked away from the less damaged area as a result of shrinkage that occurred during the charring. The severely charred material maintained some internal integrity indicating that it was a short term, intense event. The analysis indicated that chronic heating was a problem in this component. At some point the heating increased rapidly for a short period at which point the system failed.

CONCLUSION Much can be learned by examining failed components using forensic microstructural analysis. The history of the parts and the events that led to the final failure are often recorded on the surfaces and in the internal microstructure of the parts. This information is critical to understanding the nature of the failure. Explanations for failures can always be extrapolated from theory or assumptions but that is not always how things happened. An analysis of the failed parts can determine which explanations are most likely and which are inconsistent with the evidence. When decisions must be made that have a significant inherent cost it is well worth the expense to have perform a forensic materials analysis of the failure. Such an analysis not only helps characterize the failure but it can also identify other potential weak spots in the design. Forensic materials analysis and failure analysis are the subject of enough books to fill a library. These few examples are an indication of some of the types of things that can be determined. If you have questions about a piece of failed hardware it's not necessary to guess about what happened. Examining the surfaces and the internal microstructure of the hardware will often provide the answer needed.

Detect and Locate Sources of Power Transformer Deterioration Using High Speed Acoustic Emission Waveform Acquisition With Location and Pattern Recognition Ronnie K. Miller, Physical Acoustics Corporation, Princeton Junction, New Jersey Arturo Núñez, Quality Services Laboratories Plus, Trainer, Pennsylvania 1.0 Introduction An EPRI Tailored Collaboration project (R&D Agreement NO. EP-P6164/C3118) was formed to address the development of new Acoustic Emission (AE) techniques for detecting and locating active gassing sources in power transformers. This effort was initiated by EPRI along with participating utilities, Physical Acoustics Corporation (PAC) and Rensselaer Polytechnic Institute (RPI), in June 2001. The goal of this project is to demonstrate the sensitivity of AE to detect, locate and assess different active gassing sources in power transformers. The participating utilities provide a variety of power transformer to monitor and test (by PAC) in order to establish a database of transformers with known gassing conditions. RPI and PAC are working together to establish detection sensitivities for various gassing conditions through laboratory experiments. This paper discusses the overall testing/monitoring concept and reviews the progress made to date. Several case histories are reviewed in order to show examples of the sensitivity and benefits of the AE technique.

2.0 Background AE has been used in the past to test and monitor power transformers with the belief that only partial discharge (PD) and/or arcing can be detected. However, AE activity has been detected from units without PD and this evidence suggests that other sources of AE are present. In particular is the release of different gasses and the subsequent generation of gas bubbles that collapse and generate sound. We have also seen AE generated by the minute tracking in paper insulation as it breaks down and degrades. In order to investigate this behavior and properly characterize the AE response, two series of tests have been devised. First is to monitor power transformers with known gassing conditions and then compare the results with follow up internal inspection. In order to do this, each transformer is instrumented with multiple sensors and monitored for an extended period of time. At a minimum, we prefer to monitor transformer over a 12 to 24 hour period so that we can observe the effects of load swings and other operating characteristics on our data.

Besides monitoring for AE, we also collect data from other sensors such as dissolved gas oil monitors when already installed on the transformers or from periodic oil samples, current transformers, RTD’s, ambient noise sensor, AE sensors in OLTC and a zero crossing detector.

Dissolved gas oil monitors: Dissolved gas oil monitors or periodic oil samples give us an indication of the condition of the transformer based on combustible and non-combustible gas concentrations (ppm). There are several diagnostic techniques based on individual gas concentrations or ratios that allow us to detect some incipient failures on the transformer generally characterized as overheating, partial discharge, arcing or a combination of these mechanism.

Current transformers: Current transformers are used to monitor load current and pump operation. Load current monitoring allows us to track the loading peaks on the transformer. It had also been noticed that for some transformer designs, pump operation produce significant acoustic activity in areas surrounding the pumps output, therefore it is important to know if the pumps where either running or quiet and when that happened. Additionally, problems such as blocked bearing or control system malfunction can be detected Temerpature Sensors: Overheating can result in a transformer failure; therefore it is recommended to monitor the thermal behavior of the transformer. This is achieved by using magnetic RTD’s attached to transformer’s walls. Zero crossing detector: This sensor allow us to have a correlation with a 60 Hz cycle waveform electronically generated. This information will help us to identify the AE activity obtained during the positive and negative peaks of the 60 Hz signal, which is generally related to an electrical origin. Ambient noise sensor: Ambient noise is recorded during the AE test using an acoustic sensor. This information will help us to differentiate between internal and external noise. AE activity in OLTC: Acoustic sensors installed on the OLTC compartment will detect any acoustic activity produced by degradation of any insulating component and or arcing on the tap changer mechanism. The correlation of these signals with AE activity and between each other will allow us to understand the origin of the acoustic activity and disregard all the data coming from sources different than an incipient failure on the transformer.

The second series of tests are to be performed at RPI. Two different experimental apparatus are being made available in order to simulate gassing conditions in order to evaluate (and possibly quantify) the sensitivity of AE towards detecting this phenomenon. The smaller of the two apparatus allows visually observation and confirmation of the gassing process while the other is larger in scale but allows gas analysis oil as well as degassing. The results of these efforts are not reported in this paper.

3.0 Test Procedures Test procedures adopted from PAC (TP-106) are used for instrumenting each transformer and setting up the AE instrumentation. A typical layout for a three-phase transformer is shown in Figure 3.1. Bushings

6 4 16

2 3

15

5

8

10

6

12

4

14

2

1

16 Sensors Top view looking down

12 Figure 3.1 Three-Phase transformer layout. As a minimum, we start by mounting about six sensors on the high voltage side and repeat this same layout on the low voltage side. Then at least two sensors are mounted on each end of the transformer. This adds up to a minimum of sixteen sensors per transformers. Depending on the size, shape and design of the transformer, we may use up to as many as twenty four sensor to test one transformer. In addition, we utilize other sensors to provide a signal that is indicative of pump current, load current, voltage in terminals, fan on/off indication, temperatures, and hydran values. The output of the sensors monitoring these operating parameters is usually conditioned to a 0 to 10 volt D.C. signal that can be recorded on the AE instrumentation. We call this parametric data.

Conventional AE hit data and waveforms are recorded during long term monitoring. (The parametric data is recorded in two ways. First is whenever AE is detected and second is on a periodic basis in case no AE is generated.) The period of data collection can vary from twelve to twenty four hours depending on what type of load swings are experienced by the transformer and the amount of data that has been recorded after the first twelve hours of testing. For this project, we typically complete the sensor setup and calibration in the morning and have the AE instrumentation collecting data by either lunchtime or early afternoon. The data collection continues through the night until the test crew returns the following morning. If sufficient data has been collected after twelve hours, the test is terminated. Otherwise it is continued until twenty four hours have elapsed.

4.0 Case Histories

4.1 - BUTLER A core form transformer, 138 kV, 33.6 MVA, 3 Φ (see Figure 4.1.1), has had some acetylene present since October 1999. The last DGA value available (10/08/2001) indicated that this unit was gassing. The analysis (with 4 different methods) indicated the existence of discharges of low energy and arcing.

Figure 4.1.15 Transformer shown from the High Voltage side.

An AE test was performed on this unit using the sensor layout presented in Figure 4.1.2. After 20 hours and 38 minutes of continuous monitoring, 2 acoustic sources were obtained as can be seen on Figure 4.1.3. Load was switched from another transformer on the substation to this unit. The switching operation started after 15 minutes from the beginning of the test and the load was removed after 2 hours and 55 minutes. Total time with load was 2 hours and 40 minutes. As can be seen on Figure 4.1.4, the occurrence of locatable events on the transformer started after 2 hours and 44 minutes and ended after 3 hours and 8 minutes from the beginning of the test. This means that they appear after 2 hours 29 minutes and ended after 2 hours and 57 minutes from the time when the load was switched.

Figure 4.1.16 AE sensor location with sensors 1 through 4 on HV side.

CLUSTER 1

CLUSTER 2

Figure 4.1.17 3-D location plot of the AE events obtained after 20.63 hours of continuous monitoring The events occurred when the unit was loaded around 50 % of its nominal capacity. This was confirmed by the fact that 2 minutes after removing the load, the event activity ceased. Due to the location of Cluster 1, AE activity can be said to originate on the core or somewhere around the NLTC selector (connection leads). This source was active only when the load on the transformer was above 50% of its nominal capacity. The average location of Cluster 2 did not provide a clear indication of its origin. Typical waveforms recorded from Cluster 1 are shown in Figure 4.1.5.

LOAD SWITCHING PERIOD

Figure 4.1.18 Events, duration, energy and amplitude versus time graphs.

Figure 4.1.19 AE waveforms for an AE events detected on Cluster 1. (Channels 18, 16, 2, and 17 shown.) In order to get a better indication about the origin of these two clusters, it was recommended to compare the average location provided with internal construction drawings or pictures and perform an internal inspection. The results of the internal inspection are not yet available. 4.2 - WHITE VALLEY DGA values from a 3Φ core form transformer, 138 kV, 22.4 MVA (Figure 4.2.1) indicated the existence of an incipient thermal failure when analyzed with four different methods. An AE test was performed in this unit using the sensor layout presented in Figure 4.2.2. After 20 hours of continuous monitoring, 2 AE sources were obtained as shown in Figure 4.2.3.

Figure 4.2.1 Transformer shown from the High Voltage side. Two acoustic sources were detected after the AE test on this transformer. One was indicated by a cluster of events (Cluster 1) and the second by the continuous activity registered by two sensors. Cluster 1 indicates the location of an acoustic source but it was difficult to indicate an origin for this activity. Therefore, it was recommended to correlate the location provided with internal construction drawings or pictures and perform an internal inspection in order to have a better idea of its origin. Acoustic activity related to Cluster 1 was only present for 42 minutes.

Figure 4.2.2 AE sensor location with sensors 1 through 4 on the HV side.

CLUSTER 1

Figure 4.2.3 3-D location plot of AE events generated within the transformer.

The second acoustic source was detected because during the whole monitoring period, sensor 7 and 5 registered continuous acoustic activity. This activity was mostly below 50 dB and changed its generation rate along with OLTC operations. Some defects had been found in this area in similar units. This section is where the connections between winding leads and OLTC terminals are located. The acoustic activity can be due to a loose connection, PD or arcing at any terminal. Figure 4.2.4 presents a typical waveform acquired from sensor 7 during this test.

Figure 4.2.4 A typical AE waveform obtained from sensor 7 in the area near the connections between the winding leads and the OLTC.

4.3 - SEWICKLEY This core form transformer, 138 kV, 20 MVA, 3 Φ (see Figure 4.3.1) was not gassing in year 2000. When two oil samples were analyzed in 2001 (before and after the AE test) it showed the existence of ethylene (absent back in 2000) and an increase in CO2, ethane and propane. An analysis of these values with different methods indicated the existence of a thermal failure.

Figure 4.3.1 Transformer shown from the HV side. An AE test was performed in this unit using the sensor layout presented in Figure 4.3.2. After 17 hours and 42 minutes of continuous monitoring, one acoustic source was obtained as can be seen in Figure 4.3.3.

Figure 4.3.2 AE sensor location with sensors 1 through 4 on HV side.

CLUSTER 1

Figure 4.3.3 3-D location plot of AE events obtained after 17.7 hours of continuous monitoring.

One acoustic source was detected after the test on this transformer. As can be seen in Figure 4.3.4, most of the events (up to 80) were present from 14:26 hours to 17:30 hours. During this period, the load ranged from 11.06 MVA to 11.74 MVA. There is a direct correlation with the decrease of events and the reduction in load. The voltage remains practically constant during this period ranging only from 12839 to 12860 Volts. One set of waveforms from one of the cluster events is shown in Figure 4.3.5. There was another period of AE event activity that started at 1:11 hours and ended at 4:20 hours. The load during this period ranges from 7.83 MVA to 6.59 MVA. The voltage ranged from 12920 to 12880 V during this period.

Figure 4.3.3 Events, duration, energy and amplitude versus time graphs. Due to the location of the event cluster, acoustic activity may be originating in either the bottom part of the H3 High Voltage Bushing or in one of the leads going into the NLTC selector. In order to obtain a better indication of its origin, it was recommended to correlate the location provided with internal construction drawings or pictures and perform an internal inspection.

Figure 4.3.5 Waveforms for one of the events detected on the cluster. (Channels 16, 14, 2, and 12 are shown.) 4.4 - YOUNGWOOD 61 A core form transformer, 138 kV, 33.6 MVA, 3 Φ (see Figure 4.4.1) experienced a relatively significant concentration of hydrogen and an analysis of two oil samples (before and after the AE test) indicated some changes between both samples, particularly in H2, CH4, CO & CO2 gases. These gases are associated with cellulose degradation (CO2 & CO) and with partial discharge & sparking (H2 & CH4). Therefore, this may indicate an increase in insulating system degradation due to the existence of both PD and thermal failures. An AE test was performed in this unit using the sensor layout presented in Figure 4.4.2. After 17 hours of continuous monitoring, three acoustic sources were obtained as can be seen in Figures 4.4.3 and 4.4.4. A significant number of these events were obtained when the load and the voltage were increased. Most of the events (1488) were generated between the 13:11 hour and 13:14 hour of the test. Load was transferred from Youngwood 62 unit into this unit at 13:11 hours. Load records show that load was increased from 5.44 MVA (13:10 hours) to 12.41 MVA (13:15 hours). This represents an increase of 23.23 % indicating a direct correlation between the increase of load and the presence of events. The voltage represents an increase of from 12779 V (13:10 hours) to 12900 V (13:15 hours).

Load was removed from this unit at 14:31 hrs and it went from 13.37 MVA (14:30 hrs) to 7.42 MVA (14:45 hrs). Voltage had a small increase from 12890 V to 12920 V. No events were registered during this period.

Figure 4.4.1 Transformer shown from the HV side.

Figure 4.4.2 AE sensor location with sensors 1 through 4 and 17 on HV side. CLUSTER 2

CLUSTER 1

Figure 4.4.3 3-D location plot of AE events generated within the transformer for the three minutes of monitoring during load and voltage change.

CLUSTER 3

Figure 4.4.4 3-D location plot of AE events generated within the transformer for 17 hours of monitoring.

Significant acoustic activity was detected in all of the sensors (during all the monitoring period) installed at the top of the transformer. AE characteristics registered from these sensors are usually related to partial discharge activity. The most active sensor was number 3, located at the high voltage side, in front of the H1 bushing and close to NLTC leads. Some waveforms collected from the AE event cluster are shown in Figure 4.4.5. Due to the location of Clusters 1 and 2, acoustic activity can be originating in either the H3 High Voltage Bushing or in one of the leads going into the NLTC selector (located between sensors 13 and 15). Cluster 3 can be originated in one of the HV leads located between bushings H1 and H2. Even when we obtained clusters of events in specific locations that cannot always be associated with a particular component on the transformer. The acoustic activity detected and the clusters of events obtained can be generated either by degradation of one or more HV leads or an incipient failure of one or more leads of the 138 kV NLTC. In order to obtain a better indication of its origin, it was recommended to correlate the locations provided with internal construction drawings or pictures and perform an internal inspection. It was also advised to keep under surveillance individual gas concentrations trending, particularly H2 and C2H2 gases.

Figure 4.4.5 Waveforms obtained from sensors at the top of the transformers. (Sensors 1, 3, 9 and 11 are presented.)

4.5 - BETHLEN A DGA analysis of two samples (before and after the AE test) of insulating oil from a core form transformer, 138 kV, 20 MVA, 3 Φ (see Figure 4.5.1), was performed indicating the existence of local overheating. An AE test was performed on this unit using the sensor layout presented in Figure 4.5.2. After 19 hours of continuous monitoring, one acoustic source was obtained as can be seen in Figure 4.5.3.

Figure 4.5.1 Transformer shown from the HV side. One acoustic source was detected during the AE test on this transformer. Events were obtained only during a one hour period (from 11:17 to 12:17 hours). During this period, there was no significant change in either the load or the secondary voltage. Location of this cluster is not easily correlated with any specific internal part. However, according to its coordinates, acoustic activity may be originated either in the core or in one of the connections of the leads coming out from the winding. Sometimes when events are detected, we are able to hear some of them (standing next to the transformer on the NLTC selector side). We have good correlation when we can audibly hear a loud noise in that area and when we obtain an event cluster within the transformer using the 3D location plot. In order to obtain a better indication of its origin, it was recommended to correlate the locations provided with internal construction drawings or pictures, and to perform an internal inspection. It was also recommended to keep this unit under surveillance for individual gas concentrations trending.

Figure 4.5.2 AE sensor location with sensors 1 through 4 on HV side.

CLUSTER 1

Figure 4.5.3 3-D location plot of AE events generated within the transformer.

4.6 - CECIL A transformer identical to the one presented in Case History 4.4 was tested using AE in order to compare the activity recorded in both units. A DGA analysis indicated normal operation in both samples taken before and after the AE test. Figure 4.6.1 presents a view of this unit and Figure 4.6.2 indicates the sensor layout used for this test.

Figure 4.6.1 Transformer shown from the HV side.

Figure 4.6.2 AE sensor location with sensors 1 through 4 on HV side. After 22 hours of AE monitoring, the 3D plot generated from the data collected is shown in Figure 4.6.3. No acoustic sources were detected in this unit.

Figure 4.6.3 3-D location plot of acoustic emission events within the transformer.

5.0 Summary and Conclusions The Case Histories shown only represent a small part of the overall database. However, these results represent typical power transformer behavior and acoustic response. The present test procedure has proven adequate for generating sufficient data while the analysis tools have proven more than capable of identifying specific locations and areas where internal inspection should be performed. As this projects continues, we expect the database to continue to grow and produce useful data that can be applied to a variety of power transformers and operating conditions. In the next phase of this project, we will be analyzing the parametric data for correlation with the AE data. Already we have seen the impact of load changes on the acoustic response. In addition, we will initiate an effort to further analyze the AE data in an attempt to build pattern recognition and signal classifier tools. The amount of effort and the results associated with this task is worth of a paper by itself. Quantification of the AE data will be tied to the RPI experiments that were briefly described.

Overview of the Transient Performance of Coils & Windings As a Function of Their Impedance Versus Frequency Characteristic Robert C. Degeneff Mark Loose Rensselaer Polytechnic Institute Troy, New York Abstract Recently, there has been substantial interest in the frequency response characteristic of transformers. It is appreciated that it is possible to use this information to benchmark or gage the change of a winding caused by aging or mechanical movement within the windings or core. This interest has been a result of the recognition of the vast amount of information that exists in the impedance versus frequency characteristic and the potential for a powerful diagnostic tool. Today, most of the information presented in the literature is the result of measures. It is felt that considerable promise of a complementary nature exists if an analytic model of the transformer is used to explore its change in frequency characteristic. This paper will present a review of the basic relationships leading to the impedance versus frequency response of a winding from an analytic point of view. These frequencies are then related to the physical response in time within the coil. This is normally ignored and can lead to substantial errors in prediction of response. Finally, the effect of winding deformation will be used to explore the anticipated effect upon the terminal frequency response observed. This paper will deal only with single-phase systems and a follow on paper will address three-phase winding characteristics and how they relate to the single-phase response. Introduction A challenge facing utility engineers today is to apply transmission and distribution equipment in a manner that will maximize the equipment’s useful life. If this goal is achieved a method to determine the actual useful life remaining in a specific component must be found. Another way to look at this is to provide some measure of the “life” taken out of the unit by normal or abnormal use. Recently, there has been substantial interest in the frequency response of a transformer’s coils and windings. It is recognized that it is possible to use this information to benchmark or evaluate the change of a transformer, for example, caused by aging or mechanical movement within the winding system. This attention has been a result of the appreciation of the vast amount of information that exists in the impedance versus frequency characteristic of a winding and the potential for a very powerful diagnostic tool. To date most of the information presented in the literature is in the form of measured frequency response. This paper will review the methods currently used in the industry to model the transient response of the winding, show how this same model can be used to

determine the frequency response, and this present and example to illustrate the potential and challenges encountered when attempting to use an analytic model to predict the effect on frequency characteristics of a transformers response. Background Normally a transformer operates under steady state voltage excitation. Occasionally a transformer (in fact all electrical equipment) experiences dynamic and/or transient overvoltages. Often, it is these infrequent transient voltages that establish design constraints for the transformers insulation system. These constraints often have far reaching effect on the overall equipment design and cost. Obviously, the transformer must be configured to withstand any abnormal voltages covered in the design specification and realistically expected in service. Addressing the issue of transient voltage performance can be divided into three activities: recognition, prediction, and mitigation. The first is to appreciate that transient voltages excitation can produce equipment responses different than one would anticipate at first glance. For example the addition of more insulation around a conductor may in fact make the voltage stress during the transient worse and the insulation integrity of the design weaker. The transient voltage distribution is a function of the applied voltage excitation and the shape and material content of the winding being excited. The capability of the winding to withstand the transient voltage is a function of the specific winding shape, the materials voltage versus time characteristic, the past history of the structure, and the statistical nature of the structures voltage withstand characteristic. The second activity is to assess or predict the transient voltage within the coil or winding. Today this generally accomplished using a lumped parameter model of the winding structure and some form of computer solution method, which allows the internal transient response of the winding to be computed. Once this voltage distribution is known its effect on the insulation structure can be computed with a two or three-dimensional FEM. The resultant voltages, stresses, and creeps are examined in light of the known material and geometrical capability of the system in light of desired performance margins. This same lumped parameter model can be used to predict the transformers impedance versus frequency characteristic and its amplification factor. The third activity is to establish a transformer structure or configuration that in light of the anticipated transient voltage excitation and material capability, variability, and statistics nature, will provide acceptable performance margins. Of interest in this paper is the change in the performance characteristic or impedance versus frequency characteristic due to the application history of the transformer. Lumped Parameter Model A transformers (or for that matter any devices) transient response is a result of the flow of energy between the distributed electrostatic and electromagnetic

characteristics of the structure. For all practical winding, this interaction is quite complex and can only be realistically investigated by constructing a detailed lumped parameter model of the winding structure and carrying out some form of numerical solution for the transient voltage response. The most common approach is to subdivide the winding into a number of segments (or groups of turns). The method of subdividing the winding can be complex and if not addressed carefully affects the accuracy of the resultant model. The resultant lumped parameter model is composed of inductances, capacitance, and losses. This paper cannot adequately address the construction of the inductance, capacitance, and loss model for the transformer. References [7]-[11] discuss the inductance model. References [12]-[15] discuss the construction of the capacitance model. References [16]-[20] discuss the components of the loss model. Starting with these inductances, capacitances and resistive elements, equations reflecting the transformers transient response can be written in numerous forms. Two of the most common are the basic admittance formulation of the differential equation (shown in equation (1.1)) and the state variable formulation. The admittance formulation is given by reference [1]:

[ I ( s)] = [Y (s)][ E ( s)] =  [Γ n ] + [G ] + s [C ] [ E ( s)] 1 s



(1.1) Where the variables in equations 1.1 are:

[T ]

- Winding connection matrix

[T]t - Transpose of [T] [Y(s)] - Admittance matrix [C] - Nodal capacitance matrix [I(s)] - Laplace transform of current sources [E(s)] - Laplace transform of nodal voltages [Y(s)] - Transformer admittance matrix [Γ n ] - Inverse nodal Inductance Matrix =[T][L]-1[T]t [L] - Matrix of self and mutual inductances [G] - Conductance matrix, for resistors between nodes s - Laplace transform variable

In a linear representation of an iron core transformer, the permeability of the core is assumed constant regardless of the magnitude of the core flux. This assumption allows the inductance model to remain constant for the entire computation. Equation (1.1) is based on this assumption. Additionally, the various elements of the lumped parameter model are assumed not to be frequency dependent. Work in the last decade has addressed both the non-linear core characteristics and the frequency dependent properties of the materials. Progress has been made but their inclusion adds considerably to the models complexity and the computational difficulty encountered in the actual transient solution.

Frequency Domain Solution The linear differential equation shown in equation 1.1 can be solved either in the time domain or in the frequency domain. Since the model is linear the solution using either method will be identical. The frequency domain solution requires that the input wave be transformed into its frequency constituents, then these individual sinusoidal waves are applied individually to the transformer and the resultant voltage response through out the winding be determined (in magnitude and phase), and finally the total response in time be determined by summing the component responses at each frequency. This method has been used often when seeking the transient voltage response of a winding for insulation design. A benefit of this approach is that it allows the recognition of frequency dependent losses to be included easily. This may be important at very high frequencies. A disadvantage of this methodology is that is does not allow the modeling of time dependent switches, non-linear resistors (e.g., ZnO) or the recognition of non-linear magnetic core characteristics. This may be very important at more modest frequencies, e.g., below 500kHz. Solution in the Time Domain There are numerous methods to solve equations of the form of equation (1.1) but it has been found that when solving the resultant stiff differential equation model of a transformer a generalization of the Dommel's method [2,3] works very well. A lossless lumped parameter model of the transformer model containing n nodes has approximately n(n+1)/2 inductors and 3n capacitors. The following system of equations results:

[Yˆ ][ F (t )] = [ I (t )] − [ H (t )] with: 4 2 [ ] [G ] + [Γ n ] C + ∆t ∆t 2 where:

[Yˆ ] =

[ F (t )] = nodal integral of the voltage vector [I(t)] = nodal injected current vector

(1.2)

[H(t)] = past history current vector = integration step size ∆t As previously mentioned, the lumped parameter model is composed of capacitances, inductances, and losses computed from the winding geometry, permittivity of the insulation structure, permeability of the iron core and the total number of sections into which the winding is divided. Then the matrix [Yˆ ] is computed using the integration step size, ∆t . At every time step the above system of equations is solved for the unknowns in the integral of the voltage vector and then the unknown nodal voltages, E(t), are calculated by taking the derivative of [F(t)]. ∆t is selected based on the detail of the model and the highest resonant frequency of interest. Normally, ∆t is smaller that one-tenth the period of this frequency.

Accuracy Versus Complexity Any model of a physical system is an approximation. Even the simplest physical transformer has a complex winding and core structure and possesses an infinite number of resonant frequencies. A lumped parameter model, or for that matter, any model is at best an approximation of the actual device of interest. A lumped parameter model contains a structure of inductors, capacitors, and resistors which produce a frequency characteristic that contains the same number of resonant frequencies as nodes in the model. The transient behavior of this linear circuit (the lumped parameter model) is determined by the location of the poles and zeros of its terminal impedance characteristic. It follows then that a detailed transformer model must posses two independent characteristics to faithfully reproduce the transient behavior of the actual equipment in the bandwidth of interest. First, accurate values of R, L, and C, must reflect the transformer geometry. This fact is well appreciated and documented. Second, the transformer must be modeled with sufficient detail to address the bandwidth of the applied waveshape or the frequency of interest. This fact is often overlooked in practical computations. In a model valid for a given bandwidth, the highest frequency

of interest would have a period at least ten times larger than the travel time of the largest winding segment in the model. If this second characteristic is overlooked, a model can produce results, which appear valid, but may have little physical basis. A final issue is the manner in which the transformer structure is subdivided. If care is not taken the manner in which the model is constructed will itself introduce significant errors and the computation will be mathematically robust but an inaccurate approximation of the physically reality. This is an issue similar to that of meshing in a FEM calculation. Resonant Frequency Characteristic The steady state and transient behavior of any model, for any applied voltage, is established by the location of the poles and zeros of the impedance characteristic of the lumped parameter model. The zeros of the terminal impedance function coincide with the natural frequencies of the model, by definition. McNutt in reference [4] defines terminal resonance as the terminal current maximum and a terminal impedance minimum. In a physical system there are an infinite number of resonances. In a lumped parameter model of a system, these are as many resonances as nodes in the model (or the order of the system). Terminal resonance is also referred to as series resonance [5, 6]. Terminal anti-resonance is defined as a terminal current minimum and a terminal impedance maximum [4]. This is also referred to as parallel resonance [5, 6]. McNutt defines internal resonance as an internal voltage maximum and internal anti-resonance as an internal voltage minimum. The terminal resonances for a system can be determined by taking the square root of the eigenvalues of the system matrix, [A], shown in the state variable representation for the system shown: [q&] = [ A][q ] + [ B][u ] [ y ] = [C ][q] + [ D][u ]

where : [A] = State matrix [B] = Input matrix [C] = Output matrix (not the capacitance matrix) [D] = Direct transmission matrix [q] = Vector of state variables for the system & = First derivative of [q] [q] [u] = Vector of input variables [y] = Vector of output variables

(1.3)

In light of the previous definitions terminal resonance may be defined as occurring when the reactive component of the terminal impedance is zero. Equivalently, terminal resonance occurs when the imaginary component of the quotient of the terminal voltage divided the injected terminal current is zero. Recalling that in the Laplace domain that s is equivalent to jω with a system containing n nodes with the excited terminal node j one can rewrite equation (1.1)to obtain:  e1 ( s )   Z1 j ( s )  e ( s)   Z ( s )   2   2j   −   −   [i j ( s )]  = e j ( s )   Z jj ( s )   −   −      en ( s )   Z nj ( s ) 

(1.4)

The voltage at the primary (node j) is in operational form. Rearranging the terminal impedance is given by: Z t (ω ) = Z jj ( jω ) = Z jj ( s ) =

e j ( s) i j (s)

(1.5)

In these equations the unknown quantities are the voltage vector and the frequency. It is a simple matter to assume a frequency and solve for the corresponding voltage vector. Solving equation (1.5) over a range of frequencies results in the well-known impedance versus frequency plot. The amplification factor or gain function is defined as: [ N lm, j ] = =

Voltage between points l and m at frequency ω Voltage applied at input node j at frequency ω Z lj ( jω ) − Z mj ( jω )

(1.6)

Z jj ( jω )

It is a simple matter to assume a frequency and solve for the corresponding voltage versus frequency vector. If one is interested in the voltage distribution within a coil at one of the resonant frequencies, this can be found from the eigenvectors of the coil at the frequency of interest. If one is interested in the distribution at any other frequency equation (1.6) can be utilized.

Fig. 1. Input impedance of the HV winding model composed of 48 disc sections. Each section simulates one double interleaved disc. Input Impedance of the HV winding The following example uses a lumped parameter model of a hypothetical two winding single-phase transformer to illustrate the issued discussed. The winding configurations are typical but not specific to a single manufacturer. The transformer modeled has an interlaced disc high voltage winding and a helical low voltage winding. The HV winding has 96 discs with each containing 20 turns interleaved according to Chadwick/Stearns arrangement [20]. Discs were wound with rectangular wire 12.7 mm by 5.08 mm insulated with paper wrap of 1.83 mm thickness. The LV winding has 87 turns. The lumped parameter model of the HV winding was contained of 48 discrete sections with each representing a double-discs. The LV winding was represented with 10 equal winding divisions or sections. The simulation program converts the winding geometry and material characteristics into series and shunt (or parallel) capacitances, and into self and mutual inductances. The double-disc series capacitance is determined using an equivalent energy method [14], and the shunt capacitance between segments is determined with Farr’s method. The self and mutual inductance is computed with White’s method for iron-core systems [8], and closed-form expressions from Grover [21] for air-core coils.

Figure 1 is the input impedance looking into the high voltage winding of the transformer with the h

Fig. 2. Neutral Current Divided by Applied Voltage The programs accuracy has been verified over a number of years by comparison with measured winding responses. Figure 1 is a plot of the input impedance looking into the high voltage winding of the transformer with the secondary solidly grounded. Noted that in this computation the model contains 58 nodes (48 in the HV and 10 within the LV winding). A reasonably accurate representation of the poles and zeros can be expected up to the 20th resonance, e.g., approximately 250 kHz. In Figure 1 the upper plot shows the impedance to 100kHz and the lower illustrates the it to 5MHz. Figure 2 presents the neutral current divided by the applied voltage for this same transformer model. Poles and zeroes and amplification factors or transfer functions computed above 250kHz are mathematically robust but should physically be viewed with skepticism. A single double-disc wound in the Chadwick/Stearns interleaved arrangement from the HV winding was modeled on a turn-to-turn basis. The simulation program converted the double-disc physical parameters into series and shunt capacitances, as well as self and mutual inductances. The double-disc impedance was calculated and plotted against frequency up to 5 MHz, as shown in Fig. 3. In this case, the

model contains 40 nodes. The poles and zeroes are accurately represented up to approximately the 15th resonance, i.e. in excess of 5 MHz. The model used was lossless and one would expect some level of damping at these frequencies. In reality, skin and proximity effect increase the wire resistance and contribute to an increased damping of resonances above approximately 1 MHz.

b

Fig. 3. Input impedance of one double-disc modeled on the turn-to-turn basis for the Chadwick/Stearns interleaved-disc.

Fig. 4. Comparison of Input Impedance of Twin Interleaved Section, Nominal and With 0.25” Radial Bulge

Fig 5. Comparison of Input Impedance of Twin Interleaved Section, Nominal and With Shorted Turn

Fig 6. Gain Comparison At Inside Crossover –Normal Arrangement and Shorted Turn Figure. 4 provides a comparison of the nominal arrangement of a twin section interleaved disk and a winding that has experienced some event that has increased its radial build by an additional 0.25 inches. This could represent the movement one might anticipate due to the outward radial force seen in an outer high voltage winding during short circuit. In this example the variation is very small and would be difficult to use in as diagnostic tool. Not shown but a similar comparison of a twin section compressed, say from an axial force, showed a more pronounced variation. Additionally, if the winding were wound with a different disc arrangement the variation may be more pronounced. Figure 5 uses the same disk section pair but in this example one of the turns is shorted. In this example the variation between the input impedance is substantial and quite easy to see. Figure 6 is the amplification at the inside cross over for the nominal case compared to the case with the shorted turn. This also provides a very visible indication that a major change to the winding has taken place. What is missing at this point is the mechanism to use this visible change in the impedance characteristic to predict the reason for the change in impedance or gain or transfer function.

Fig 7. Comparison of Input Impedance of Twin Disk Sections – Continuous and Interleaved (Chadwick/Stearns & Van Nuys) Figure 7 simply compares the impedance versus frequency of three different disc winding arrangements – the continuous and the interleaved (interlaced) – using the Chadwick/Stearns and the Van Huys arrangements. The series capacitance of both interleaved winding arrangements are the same and as such the response shown in Figure 1 would be representative of either winding type. Clearly, then it is important to represent the winding at the turn level if the response above 500kHz is of interest.

Conclusion and Discussion The lumped parameter representation of a transformer modeling disc section pairs groups of turns is an adequate representation when computing the transient voltage response for standard factory impulse tests (full wave, chopped wave, switching surge). The impedance versus frequency and transfer function is shown in Figure 1 and 2. For this model the upper range of validity is 250kHz and this is a representative of most lumped parameter models used for insulation design. With this level of detail it would be incorrect to use the model to predict a transformers frequency response or transfer function at higher frequencies. However, a turn-to-

turn model would provide a model with a frequency bandwidth in excess of 5MHz. This is shown in Figures 3-7. There are several advantages of an analytic model appropriate for high frequency. It can be used to explore and define the effect on impedance versus frequency and transfer function of many winding changes, e.g., aging, movement, failures. It would provide a tool to explore various winding configurations providing insight to winding performance not possible by measurements alone. Finally, it is potentially possible to use the measure response of a system to synthesize a lumped parameter model, rather than starting with the transformers physical parameters if these were not available. This would then allow variations to be explored analytically even if the physical data representing a transformer where not available. This model then could be used to assess the cause and potential seriousness of measured impedance versus frequency and/or transfer functions.

Reference 1. R.C. Degeneff, “A General Method For Determining Resonances in Transformer Windings,” IEEE Trans. on PAS, Vol. 96, No. 2, March/April 1977, pp. 423-430. 2. H.W. Dommel, “Digital Computer Solution of Electromagnetic Transients in Single and Multiphase Networks,'' IEEE Trans. on PAS, April 1969, pp. 388-399. 3. R.C. Degeneff, “Reducing Storage and Saving Computational Time with a Generalization of the Dommel (BPA) Solution Method,'' IEEE PICA Conference Proceedings, May 24-27, 1977, pp. 307-313. 4. W.J. McNutt, T.J. Blalock, R.A. Hinton, "Response of Transformer Windings to System Transient Voltages," IEEE PES Transactions, Vol , 1974, pp. 457-467 5. P.A. Abetti, “Correlation of Forced and Free Oscillations of Coils and Wingings,'' IEEE PAS, December 1959, pp. 986-996 6. P.A. Abetti, F.J. Maginniss, ``Fundamental Oscillations of Coils and Windings,'' IEEE PAS, February 1954, pp. 1-10 7. L. Rabins, “Transformer Reactance Calculations with Digital Computers,” AIEE Trans., Vol. 75, Pt. 1, July 1956, pp. 261-267. 8. W.N. White, An Examination of Core Steel Eddy Current Reaction effect on Transformer Transient Oscillatory Phenomena, General Electric Technical Information Series, No. 77PTD012, April 1977. 9. R.C. Degeneff, ``A Method for Constructing Terminal Models for Single-Phase n-Winding Transformers,'' IEEE Paper A78 539-9, Summer Power Meeting, Los Angles, 1978 10. R.C. Degeneff, W.N. Kennedy, ``Calculation of Initial, Pseudo-Final, and Final Voltage Distributions in Coils Using Matrix Techniques,'' Paper A75-416-8, Summer Power Meeting, San Fransico, CA 11. Frank M. Clark, Insulating Materials for Design and Engineering Practice, New York, Wiley, 1962. 12. Von Hippel, Dielectric Materials and Applications, MIT,1954.

13. R.C. Degeneff, Simplified Formulas to Calculate Equivalent Series Capacitances for Groups of Disk Winding sections, General Electric, TIS 75PTD017, August 16,1976. 14. A. Scheich, ``Behavior of Partially InterleaveTransformer Windings Subject to Impulse Voltages,'' Bulletin Oerlikon, No. 389/390, pp.41-52 15. J. Lammeraner and M. Stafl, Eddy Currents, The Chemical Rubber Co. Press, Cleveland,1966. 16. F. de Leon and A. Symlyen, Complete Transformer Model for Electromagnetic Transients, IEEE Transactions on Power Delivery, Vol. 9, No. 1, January 1994, pp. 231-239. 17. E.J. Tarasiewicz, A.S. Morched, A. Narang and E.P. Dick, Frequency Dependent Eddy Current Models for the Nonlinear Iron Cores, IEEE Trans. on PAS, Vol.8, No.2, May 1993, pp. 588-597. 18. J. Avila-Rosales and Fernando L. Alvarado, Nonlinear Frequency Dependent Transformer Model for Electromagnetic Transient Studies in Power Systems, IEEE Trans. on PAS Vol. PAS-101, No. 11, Nov. 1982. 19. R. Batruni, R. Degeneff, M. Lebow, ``Determining the Effect of Thermal Loading on the Remaining Useful Life of a Power Transformer from Its Impedance Versus Frequency Characteristic,'' IEEE Transactions on Power Delivery, Vol 11, No. 3, July 1996, pp 1385-1390 20. A.T., Chadwick,, J. M., Ferguson, , D.H., Ryder G.F., Stearn, “Design of Power Transformers to Withstand Surge Due to Lightning with Special Reference to a New Type of Winding”, (IEE Proc. Part II, Vol. 97, 1950, p. 737) 21 F.W., Grover, “Inductance Calculations”, (Dover Publications, Inc., N.Y., 1980)

Gasses Generated From Bare Electrodes at Elevated Temperatures Immersed in Mineral Oil Observed over Ranges of Temperature and Pressure Ron Miller, Barry Ward, Ahmad Shahsiah, Robert Degeneff Abstract Invaluable data about transformer performance and performance trends can be obtained by effective analysis of dissolved gases in transformer oil as well as analysis of the gas in transformer headspace. One method of determining the gasses present is to record acoustically the frequency of the gas bubble generation. Gas chromatography is another popular method used for this type of analysis. Fault conditions such as overheating, partial discharge, and arcing produce gas in oil. It is appreciated that different gasses and rates of production are generated at different temperatures and pressures however, there is little known about the rate and time constant of generation of such gases. This paper presents the data measured in a test cell at RPI (Rensselaer Polytechnic Institute) where a bare probe was heated to 200 degrees C while immersed in mineral oil with a head pressure between 0 and 10 psi. The gas in oil components and trend of gas generation was determined by taking repeated gas in oil and gas in headspace sample and analyzing with a Shimadzu gas chromatograph. The rate of gasses generated is closely monitored after the oil experienced an impulse of temperature for a short period of time, which mimics a hotspot in a power transformer. The amount and components of generated gases are presented.

1. Introduction Electric utilities are under pressure to both extend the life of their installed equipment and to load that equipment beyond its normal design limit. Over the past decade substantial efforts have been expended in an effort to achieve both of these goals. Invaluable data about transformer performance and performance trends can be obtained by effective analysis of dissolved gases in transformer oil as well as analysis of the gas in transformer headspace. Gas chromatography is a well-accepted method used for this type of analysis. Another method of determining the gasses present is to record acoustically the frequency of the gas bubble generation. Fault conditions such as overheating, partial discharge, and arcing produce gas in oil. It is appreciated that different gasses and rates of production are generated at different temperatures and pressures however, there is little known about the rate and time constant of generation of such gases. Additionally, transformers rarely operate at equilibrium and any insight obtained into the condition of the transformer must understand not only the gas generation method but also the thermal and chemical dynamics of the transformer environment. This paper is a preliminary step in establishing the rate and gases generated from a bare electrode immersed in oil under a variety of conditions.

This paper presents the data measured in a test cell at RPI (Rensselaer Polytechnic Institute) where a bare probe was heated to 200 degrees C while immersed in mineral oil with a head pressure between 0 and 10 psi. The gas in oil components and trend of gas generation was determined by taking repeated gas in oil and gas in headspace sample and analyzing with a Shimadzu gas chromatograph. The rate of gasses generated is closely monitored after the oil experienced an impulse of temperature for a short period of time, which mimics a hotspot in a power transformer. The amount and components of generated gases are presented. The experiment was then repeated with acoustic monitoring apparatus and the frequency characteristic recorded and compared to the known gases generated as determined by gas chromatograph.

2.Test apparatus The test vessel in which the experiment was conducted is made of iron (similar to a transformer tank construction) 12’’ by 12’’ by 18” with the interior coated with an epoxy resin manufactured by Hentzen Coatings Inc. The product is known as epoxy catalyst 05599CEH. Test vessel provides a closed and controllable environment for the transformer oil for the purpose of this experiment. The test vessel has viewing ports on the sides for visual inspection of the oil. The vessel has a sufficient number of inlet and outlet ports to allow transfer of degassed oil into and samples of gas and oil out of the vessel. Additionally, provisions are made for mounting the heater to generate the hot spot, and special

path for inert gas and vacuum pump. Figure 1 shows the schematic of the test system. It consists of the following major components:

Test vessel and configuration: Consists of a tank with two sight glasses so that the heating element immersed in oil is visible. The tank contains about 4 gallons of oil. The heating element and thermocouple attached to the surface of the heater. Heater: The heater is a 5000watt cartridge heater manufactured by Watlow, known as Firerod. The heater is a rod diameter is 5/8”, single phase 480 Volts power input, and watt density of 300 watt/in2. Watt density of the heater is important because it is desired to create a hot spot in the oil rather than heat up the entire volume of oil. Even though there is no forced circulation of oil inside the test vessel, the surface heat of the heater is quickly dissipated by the natural convection of oil. Therefore the heater must have a great enough Watt density. Prior to construction of the test system, heat transfer calculations were done to determine the required Watt density of the heater prior to placing an order for one. These calculations are based on Russelt and Reyligh numbers and are based on the calculations of natural convection heat transfer inside the oil [1]. The calculations were based on creating a hotspot up to 400 Degrees C and the result was Watt density of 380 Watts/in2. The temperature controller: The used controller is Watlow series F4 ¼ DIN industrial ramping controller. This PID controller provides a control signal to a SCR, which limits the power into the heater from a 150kVA step-up transformer. This step up transformer is the main power source of the heater for the entire thermal loop. Of course because the heater Wattage is smaller than the source, the controller should be set properly to limit the maximum allowable energy to pass to prevent damage to the heater. Controller specifications are: calibration accuracy and sensor conformity for T type thermocouple: 0.1% of span +/- LSD 25+/-3 degrees C ambient and accuracy span less than or equal to operating ranges

The oil type: The insulating oil used for the experiment is Shell Diala-A. It is a Shell product and meets standard specifications required by both ANSI/ASTM D3487 and NEMA TR-P8-1975for domestic electrical oil applications. These products have high electrical resistance and are thermally and oxidatively stable.

Vacuum/pressure gauge

Transformer Oil

Temperature Controller Vacuum pump

Gas

Inert gas

Adjustable fixture Heater & Thermocouple Window

Window

Oil

Figure1 Schematic diagram of the test vessel

Degassing system: the degassing system is used to purge the system from gas and moisture. Figure 2 is a schematic of degassing system. The bypass loops and the necessary valves make it possible to purge gas and moisture from main loop, test vessel, or the two systems connected together. Degassing process is done by spraying the oil to the walls of the degassing chamber and therefore making a thin film of oil on its wall and at the same time vacuuming the degassing chamber. This causes the dissolved gasses and moisture in oil to be purged from the thin oil layer and therefore makes the degassing process faster. The pump in this system is a 3phase 5Hp pump and a frequency variable speed controller controls the pump speed.

To the main loop Degassing Chamber

Test vessel Vacuum gauge

Vacuum pump Pump

Figure2 Schematics of degassing system

3. Description of Experiment The goal of this experiment is initially to establish the power and duration of heating to create a rate of gas generation from a bare electrode immersed in oil at various temperatures and headspace pressure that can be meaningfully observed with a gas chromatograph. The next step is to repeat the experiment and observe gasses and trends of gas generation with an acoustical monitoring system. Simultaneously, gas in oil will be measured with the gas chromatograph. To monitor the gas contents of oil Shimadzu Togas system is used. The method of the experiment is as below: a) Preparation of oil -

-

-

The test vessel is first evacuated to 1.5*10-2 torr (McLeog gauge) using a Sargent-Welch high vacuum pump. Then the system is filled with Nitrogen to a positive pressure of 2 psi above atmospheric pressure. The degassing chamber, which is normally under vacuum to degas the oil, is also brought back to atmospheric pressure by injecting Nitrogen. The valve separating the two systems is opened letting the clean degassed oil inside degassing system to flow into test vessel due to pressure difference of the two systems. The pump in the degassing system is used to force the oil into the test vessel if the oil level inside the test vessel has not reach the required amount (approximately 4 gallons inside the test vessel) due to pressure difference. The connecting valve between the two systems is closed, the pressure inside the tank is set to the predefined value by venting the gas or injecting more Nitrogen The controller is set to keep the oil temperature at 35 degrees C as the base temperature, and The system is left to reach equilibrium for at least 12 hours.

b) The Method and Time Intervals: The results of two preliminary experiments will be presented to establish a base of gases and generation rates produced at 200 degrees C and headspace pressures of zero and 10 psi. After the system reaches equilibrium six samples are taken, three oil and three gases. At a given time one oil and one gas sample is taken and then after a 15 minute interval the next two samples are taken, and after another fifteen minutes the final two samples are taken. Analysis of these samples takes approximately 3 hours. At the end of this analysis period a heating impulse of 200 degrees C for a period of 3 minutes is applied to the system. See Figure 3. Immediately after the oil is back to its equilibrium at the base temperature (35 degrees C) another six samples are taken using the same data pattern used prior to the heating impulse. Analysis requires an additional period of approximately three hours. Once this is accomplished another six samples. This scenario is repeated again after at least 12 hours. The above method gives 27 data points for each oil and headspace gas. At the end the average of each three data sets for gas and oil samples are calculated and 9 data points for the total gas and oil analysis are obtained. This whole process is repeated once with atmospheric pressure inside the test vessel and once for the headspace pressure of 10 psi. Figure 3 shows the applied temperature impulse pattern.

200 Degrees C

35 Degrees C

1’

3’

10’

Figure 3 Temperature impulse pattern 3. Gas chromatography results There were two experiments conducted as described in the previous section. In the first experiment the headspace pressure was 2psi. This small positive pressure in order to assure that no air would flow into the system due to system for any reason. Figure 4 shows the timing of the first experiment and the resulted change in oil and headspace gas contents are shown in Figures 5 and 6 respectively.

0

18.75

23

42

: Three measured data sets : Data generated based on the slope of changes : Heating impulse as shown in Figure 3

Figure 4 Time line diagram of the first experiment

45

50.25

Time Hours

Gas in oil concentration (2psi) 200 H2 CH4 CO CO2 C2H4 C2H6

ppm

150 100 50 0 -50

0

20

40

60

hours Figure 5 Gas in oil concentration in first experiment

Gas concentration in headspace (2psi) 40 H2 CH4 CO CO2 C2H6

ppm

30 20 10 0 0

20

40

60

hours Figure 6 Headspace gas concentration in experiment1 In Figure 5 and 6 in order to show the change in gas concentration just prior to applying the heating impulse, as there was no measured data right prior to the impulse, a data point was computed based on the slope of the change of gas concentration at times 18.75 and 45 hours. In the second experiment the headspace pressure was increased to 10 psi and the same pattern of data extraction was followed using a new degassed oil sample. Figure 7 shows the timing of this experiment and the results of oil and headspace gas concentrations are shown in Figures 8 and 9.

5

8.5

24.75

28

31.5

66.5

: Three measured data sets : Data generated based on the slope of changes : Heating impulse as shown in Figure 3

Figure 7 Time line diagram of the second experiment

Gas in oil concentration (10psi) 120 H2 CH4 CO CO2 C2H4 C2H6

100 ppm

80 60 40 20 0 0

20

40

60

80

hours Figure 8 Gas in oil concentration in second experiment

Gas concentration in headspace (10psi) 50 40 ppm

0

H2 CH4 CO CO2 C2H6

30 20 10 0 -10 0

20

40

60

80

hours Figure 9 Headspace gas concentration in experiment2

Time Hours

4. Acoustic monitoring results

5. Discussion -

-

A hotspot in insulating oil even at modest temperatures causes a relatively large amount of gas to be generated. In order to be able to keep track of the trends of the generated gases a heating impulse for a short period of time was imposed to the system. This hotspot generates enough gases to study for the purpose of the experiment and yet the amount of generated gases are not too high to be immeasurable using gas chromatography method. In order to reduce the effect of noise in the measured data each time three data sets are taken and analyzed and the average of these points is considered as the reliable data. Typical variation between readings taken under the same conditions are on the order of 25%.

6. References [1] Frank P. Incropera, David P. DeWitt, “Introduction to heat transfer,” John Willy & Sons, 1996 [2] Herbert G. Erdman, ”Electrical insulating oils,” STP 998, ASTM 1988 [3] “ Basic transformer life characteristic”, Report No EL-2443, Electric Power Research Institute, 1982 [4] ANSI/IEEE C57-104-1991,” IEEE guide for detection and determination of generated gases in oilimmersed transformers and their relation to the serviceability of the equipment,” [5] ASTM D2779-1992,” Standard test method for estimation of solubility of gases in petroleum liquid,”

3 CIRCUIT BRAKER AND OTHER SUBSTATION EQUIPMENT DIAGNOSTICS Chair: Bjorn Holm, International Switchgear Consulting Ltd.

Circuit Breaker Operator Analyzer Randy Cox TXU Electric

SUMMARY For many years utilities have experienced problems with transformer outages and circuit breaker failures directly associated with circuit breaker operator issues, which have created an unfavorable environment for the utility and the utility customer associated with these incidents. Traditional testing, while valuable, was not able to identify the source of the problem. Many of these issues have been associated with the lubrication inside the breaker operator, while others have been associated with the electrical/mechanical design of the breaker. Attempting to identify the “root cause” of these problems has been extremely difficult until recently. BACKGROUND Power companies use what is typically referred to as a “DFR” (Digital Fault Recorder), and distribution operating reports for determining fault analysis on high and low voltage faults, or equipment problems. In the past, these systems, while useful, have not been able to provide the insight necessary to determine the exact location of the problem. Once a breaker had been identified as “slow” to operate, diagnostic crews were dispatched to time and test the breaker associated with the issue only to be cleared of any problems. Once the fault occurred, the breaker had then operated, providing the exercise necessary to free the mechanism. Further timing tests revealed that the breaker was within specification, leaving the Diagnostic technician at a loss as to what caused the slow trip. This is a problem that has been occurring over an extended period of time, but at a lower frequency than what we see today. Our maintenance practices in the past have allowed for a more frequent look at our circuit breaker operators, and the electrical/mechanical checkout of the breaker. Many utilities in years past had what is referred to as an “Annual Maintenance Program”. This program was designed to lubricate the operating mechanism and repair any electrical/mechanical issues, which kept the operator “free” to move when called upon to do so. In actuality, utilities were using what has for years been incorrectly referred to as “lubricants” when in fact they were penetrants. These “penetrants” came in the form of a spray can, which actually dissolved all of the lubricating oil from the grease, leaving only the soap for the operator bearings to move upon. The Lubrication Guide of the Doble Circuit Breaker Committee [1] discusses these issues of lubrication and maintenance practices at length. An excerpt taken from this guide is as follows:

Transmission and distribution breakers have become a maintenance challenge because of their extended time in the field without relubrication … Circuit breaker failures have been traced to faulty lubricants and/or questionable lubrication practices. Many circuit breakers in use today are very old and continue to be lubricated with what the manufacturer specified many years ago. Many of these lubricants will not last. Some greases will separate, leaving only a dry thickener which can slow the breaker action. Some greases and penetrating oils can change in physical form, leaving what appears to be a varnish-like residue in bearings and other critical friction areas. [1] The Lubrication Guide of the Doble Circuit Breaker Committee [1] also goes on to explain how the use of synthetic greases in lieu of petroleum-based greases should provide longer service life of the breaker operator between lubrications.

We now know that the answer to this specific problem is not found in traditional diagnostics. PREVIOUS TECHNOLOGY The idea of monitoring the trip and close operations on a circuit breaker has been around for many years. In the past such an undertaking was not only expensive, but it was also time consuming and limited in scope. The following are a few examples: • • •

These monitoring devices had to be “hard-wired” primarily in the field. The ability to locate a problem depended upon the fact that it was hardwired to a problem breaker. The cost of these units prohibited their use on an “across the board” basis (for a system of any size, prioritizing where these units were to be installed was unavoidable, running the risk of missing the problem breakers).

This technology had limited effects, as these units only triggered when the breaker saw a fault, leaving the utility with the same result, which was an outage. Part of the problem with this method of operation is that these units were typically installed on new breakers, and these new breakers had new grease in the operators. More than likely, these breakers would not develop operator problems for several years, leaving the utility with the same problem that they had prior to utilizing this technology. CURRENT TECHNOLOGY With the current demand for utilities to improve system reliability and financial return, test equipment manufacturers have had to step up to the plate with regards to creating a more “user friendly” and “cost effective” method of gaining the information necessary to determine where these breaker operator problems are located. One such piece of test equipment is the Kelman Profile P1 Circuit Breaker Analyzer. The Kelman is a lightweight, portable unit, which is very simple to use. It only takes about 10 minutes to actually run the test while the breaker is in service, and the information that is obtained has proven to be extremely valuable in identifying problems associated with breaker operators

and other issues that can cause a breaker not to perform correctly when called upon to do so. Utilizing this test equipment allows the user to make the necessary arrangements via switching in case of a problem operator, eliminating the risk of a customer outage. The following is a brief list of the typical information that can be obtained by the Kelman. • • • • • • • •

D.C. voltage supply (station battery). A.C. charger supply. Trip and close coil operating information. Initial contact make and break times. Trip latch information. Auxiliary contact information. Instantaneous relay delay time. Multiple operation overlay capability.

IMPLEMENTATION Connection In a paper written by Wesley R. Speed with TXU Electric titled, CIRCUIT BREAKER OPERATOR SIGNATURE ANALYSIS [2] a description of the connection and typical information obtained are described in detail: The connection of the breaker analyzer is extremely simple, consisting only of four leads, as indicated the Figure 1. First, two leads are attached to the DC voltage supply to the circuit breaker. Then, an AC clamp-on is connected to the secondary of one phase of the current transformers (providing main contact operating time). Finally, a DC clamp-on is connected to the trip and/or close circuit (providing DC current amplitude over time). The CT can also be placed ahead of the instantaneous relay coil, which will provide the user with the instantaneous relay delay time.

Figure 1 – Connecting The Circuit Breaker Analyzer

A typical shot obtained from the breaker analyzer is shown in Figure 2. There are three main features – the main contact time, the DC supply voltage, and the trip / close coil current during the operation of the breaker. How the mechanical movement of the operator corresponds to the electrical trace is indicated in Figure 2.

movement of the operator corresponds to the electrical trace is indicated in Figure 2. When a trip signal is initiated to the breaker, the DC current through the trip coil starts to rise, moving the plunger inside the coil. The plunger moves until it hits the trip latch of the breaker. This is indicated on the electrical trace by the valley after the first hump on the curve. Up to this point, the breaker is stationary. For a proper breaker operation, the trip latch is released, the energy in the stored energy mechanism starts in motion, and the breaker contacts start to move. At some point the main contacts break the current through the breaker, and an auxiliary contact opens and breaks the current in the trip circuit. Meanwhile, the DC supply voltage is measured during the shot.

Trip Coil Solenoid Hits Latch

Main Contact Time ‘a’ Contact Opens DC Supply to Breaker Breaker Comes off Latch, Operator Starts to Move

Figure 2 – Typical Trip Shot Initial Program

of the problems found during this first round of testing were related to “failure to re-close” while the rest were related to slow operators, failure to trip, and relay coil and contact problems. After a breaker has been idle for a long period of time, and if penetrants have been used (i.e.….past maintenance practices), or the grease is old, then the operating pins, rollers and bearings tend to stick in place preventing the breaker operator from any sudden movement. As described earlier, this can be very difficult to detect if you do not have the initial trip. Once the breaker has been operated, these rollers, pins and bearings tend to free up, allowing the breaker to operate within normal specs eliminating the possibility of finding the initial problem. Considering the fact that we were not able to immediately work every operator that tested slow, the first order of business involved lubrication utilizing some form of approved oil (not all oils are compatible with all greases). If the breaker operator then operated at a faster speed, the operator was placed on a prioritized list for operator overhaul. For years utilities have replaced operating coils in breakers under the assumption that these coils were defective, when in reality they were overworked. Many trip and close coil failures can be traced back to the fact that these coils stayed energized much longer than they were designed to, due to slow operators. The following graph will help to clarify this statement.

Figure 3 – Example of Faulty Lubrication We began the program utilizing the following guidelines: • • • •

Main Contacts < 50 ms on trip shot (3 cycle breaker) Main Contacts < 200 ms on close shot Breaker off latch < 17 ms (1 cycle for a 3 cycle breaker) Voltage drop < 10% • Instantaneous Relay Setting = 5 cycles While the benefits of the program were greater than expected, we also desired to improve the low voltage relaying performance. We knew that our breaker operators were

in trouble due to past maintenance practices, but we also knew that our control circuits and relaying schemes were causing some serious problems as well (failure to re-close…..failure to trip). At this point, we identified a Diagnostic Technician and a Senior Electrician, and trained them to trip test our low voltage system via relay, utilizing the Kelman Profile P1 Unit in the instantaneous and overcurrent relaying circuits of the operation of the breaker as shown in Figure 4. In only the first few months of the program, 410 breakers were tested. Out of that number 82 problems were found, and 52 were repaired on site.

Overload operation Instantaneous operation

Figure 4 – Instantaneous and Overload Relay Test, Same Operator Lessons Learned We learned many lessons in the implementation of the circuit breaker analysis program. These are briefly described as follows: • •

•

Data management is critical. We purchased the breaker analyzers with a bar code feature that proved to be extremely helpful in maintaining “consistent” data. While close shots will provide the user with good information, it is typically the trip circuit that causes the utility the most problems. In order to gain the insight necessary for good performance, it is crucial to obtain the first trip. As stated earlier, without this information it is difficult at best to determine where the actual issues are. At each test interval it is very important to connect the analyzer to the same points. For this reason it is important to mark the location of the test equipment connections.

• • • •

• •

We have found that there are some solenoid breaker operators that will pass a test even though the grease within the operator is failed. Eventually these issues will prove to be fatal even though they are not apparent at the time of the test. Vacuum breakers will not provide the same amount of stored energy as the circuit breakers of old. Trip and close coils need to be sized properly (we found several 125 VDC coils on 48 VDC systems). Tripping a circuit breaker via the instantaneous attachment on the overload relay can provide great insight into the cause of many “failure to re-close” problems, since this action will immediately call for a re-close (we have found that failure to re-close is one of the largest issues with circuit breaker performance). Connecting the leads of the Kelman to a circuit without first testing the circuit to be sure that it is a D.C. circuit will damage the unit. Overcrowding the data on the unit can cause you to lose information.

CONCLUSION The circuit breaker operator analyzer has proven to be the most effective tool in determining the “root cause” of circuit breaker operator problems. Past maintenance practices were only considered effective because of the time interval associated with them, when in actuality they were destructive in nature. As a result of the information that we have now gained from the Kelman, we have an annual trip-testing program in place, and have greatly improved our high and low voltage breaker performance. We have also established a baseline test on all of our breakers so that we can now see when our operators initially begin to deteriorate in performance. This method of operation has allowed us to be in a position to prioritize our work utilizing useful information.

REFERENCES [1] [2]

Doble Circuit Breaker Committee, Lubrication Subcommittee: Lubrication Guide of the Doble Circuit Breaker Committee, 1995. Wesley R. Speed, TXU Electric: CIRCUIT BREAKER OPERATOR SIGNATURE ANALYSIS

BIOGRAPHY Randy Cox is the Diagnostic Technician Supervisor at TXU Electric. His current responsibilities include; Oversight of field diagnostics in the Dallas District, Chairman of the Oil Diagnostic Committee, and Chairman of the Diagnostic Training Committee for the Transmission Business Unit.

Solutions for SF6 Leak Detection, Diagnosis, Sealing and Management

Mike Alfieri ConEd Nick Dominelli Ian Wylie Powertech Labs Inc. Luke van der Zel EPRI

Introduction This paper addresses four important aspects of SF6 gas: - Diagnosis of High Voltage Apparatus via SF6 Analysis, - Utility experience of SF6 Leak detection using the latest SF6 camera, - Utility approach to SF6 Management using MMW (Maintenance Management Workstation), - Utility experience on temporary SF6 leak sealing. SF6 has a global warming potential 24,000 times greater than CO2, the most plentiful greenhouse gas. The estimated life of SF6 in the atmosphere is 3200 years compared to CO2, which has a lifetime of 200 years. At present the US EPA (Environmental Protection Agency) is requesting users of SF6 gas to reduce the release to atmosphere through voluntary agreements. This plan is the first step towards reducing our SF6 usage as well as its release to atmosphere. At present, electrical utilities are estimated to use 80% of the SF6 manufactured worldwide.

Diagnosis of High Voltage Apparatus using SF6 Analysis Assessment of SF6 is valuable in the following instances: - determine acceptance of new gas and comply with specifications - commission new equipment to verify proper filling procedures - monitor air content following repairs and refilling operations - verify suspected faults - locate suspected faults in bus sections - assess faulted gas for continued use and for personnel safety, and diagnose equipment

The development of field assessment methods and a field detector for real-time continuous monitoring of low ppm (parts per million) levels of decomposition products are presented and discussed. New SF6 for electrical equipment is supplied with guaranteed low levels of contamination with air, moisture and carbon tetrafluoride (CF4). However, under arcing (Boudene, C., et al , 1974) or overheating conditions (Chu, F. and Massey, R., 1982), small amounts of SF6 decomposition products are produced. The specific product and amount depends on the type and intensity of the fault, the type of material involved, and the level of moisture, oxygen, and other gaseous contaminants present. Low levels of products do not adversely affect the SF6 dielectric properties. However, because most of these products are highly corrosive and very toxic, their accumulation raises concerns about equipment reliability and personnel safety during maintenance procedures and clean up of faulted equipment. (Chu, F, 1985) Regular analysis of SF6 gas can provide utilities with invaluable information essential for safe handling and proper equipment operation. Of particular interest is its potential as a diagnostic technique for in-service equipment. In principle, knowledge of the nature and relative amounts of SF6 products can be used to determine the operating condition of the equipment, to locate faults and to detect incipient faults in GIS (Gas Insulated Substations). Several attempts have been made to use SF6 analysis for equipment maintenance purposes (Mastroianni, M., 1985, Dwyer, F.J. and Miners, K.B., 1982, Tominaga, S. et al, 1981). The recommended IEC and CIGRE levels for gaseous impurities are summarized in Table 1. It is apparent from Table 1 that analysis for diagnostic purposes must have detection limits comparable to new gas. The analysis of these compounds at low parts per million (ppm) levels poses a challenge of its own.

TABLE 1 Contaminant Levels in SF6 as Recommended by IEC and CIGRE

Contaminant

IEC New gas

Air

0.05%w (2517 ppmv) 0.05%w (800 ppmv) 15ppmw (122 ppmv)

CF4 H2O SF4, WF6, SOF4, SOF2, SO2, HF, SO2F2

Criteria CIGRE Reclaimed gas

CIGRE In-Service gas

2%v

3%v

(included in air content)

(included in air content)

1600 ppmv at 100 kPa

4000 ppmv at 100 kPa 100 ppmv

1.0ppmw as HF (7ppmv as HF)

50 ppmv total 2000 ppmv

In a recent EPRI sponsored project (EPRI, 2000), existing and emerging technologies suited for comprehensive field assessment were investigated. The results indicated that a customized portable gas chromatograph was capable of analyzing mixtures of O2, N2, CF4, SOF2, SO2F2, COS, SO2, and

H2O in SF6 in a single run. Details of the research have been reported elsewhere (Dominelli, N. and Wylie, I., 1993 and 2001). Briefly, the portable unit consists of a portable GC (Gas Chromatograph), equipped with a built in high pressure manifold, sampling pump, and in line frit. It can sample directly from GIS equipment and SF6 cylinders. The entire analysis requires less than 3 minutes and uses less than one gram of gas. Field operation is straightforward and requires minimal knowledge of gas chromatography. The software prints or displays a simple report giving the results in ppmv (parts per million by volume) or %v. Ideally, on-site and real-time field detection of SF6 products would overcome the problems associated with sampling and transportation of these reactive and toxic gases. Such a system need not have the sophistication of laboratory techniques, but by focusing on the key decomposition gas thionyl fluoride (SOF2), it would allow the user to screen out the bulk of GIS equipment on-site while identifying equipment requiring additional testing. Based on results to date (Dominelli, N. and Wylie, I., 1996), such a screening tool would reduce the number of samples needing costly laboratory analysis. More importantly, it could be used to alert maintenance personnel to potential safety hazards. In response to this need, EPRI and Powertech Labs collaborated in the development of a prototype field detector for faulted SF6 gas and evaluated in the lab and recently it was further refined and field hardened. The assembled unit is shown in Figure 1. After direct connection to the electrical equipment, the detector responds almost immediately once the flow begins and requires less than one minute of constant flow at 100 mL/minute to get a representative reading. Following equilibration, the total SF6 decomposition products (mainly SOF2 and SO2) can be read directly in ppm. The major application of the DPD is to provide a quick and accurate measurement of SF6 decomposition products in field situations. It is advantageous to test the gas at the source due to the unstable nature of low-level decomposition products and to detect faults quickly without having to wait for lab analysis. With this detector, rapid screening of SF6 is possible to quickly locate problems and minimize outages. Confined spaces (with a sampling pump) and equipment compartments can be rapidly assessed for personnel safety before maintenance begins so that appropriate procedures and precautions can be implemented.

SF6 Decomposition Products Detector FIGURE 1

In summary, two valuable on-site SF6 Analysis tools have been produced through EPRI funded research, in collaboration with Powertech: - Firstly, the SF6 Decomposition Products Detector. This small, portable detector allows rapid screening of SF6 in the field, - Research is in-progress on the application of the tailored portable GC – which will provide a onestop answer on the entire spectrum of components needed in an analysis.

Utility experience of SF6 Leak detection using the latest SF6 camera In a six-month trial period Consolidated Edison used the new generation GasVue TG-30 LIS (Laser Imaging Systems) camera extensively. It was applied to search for leaks from indoor and outdoor equipment in a variety of weather and temperature conditions. No shipment by common carrier occurred following initial shipment. During this trial period the camera was operated for over 200 hours. No down time was recorded and the camera was never sent to the shop for repair. This reliability record is significantly better than the experience EPRI has had with its two laser cameras from the first generation GasVue. It should be added that the EPRI first generation cameras are regularly shipped by common carrier for use at many utilities and this may also be a factor in their reliability. Although calibrated tests were not conducted, Consolidate Edison reported that the ability of the prototype to detect leaks seemed excellent and probably better than the earlier version, which they had also used. We have applied the camera to a variety of their substations and were able to locate a significant number of leaks. Consolidated Edison carefully evaluated the operation of the improved camera prototype and reported several design details, which they felt, could be improved. LIS responded by considering the report and acting on those they felt were both useful and achievable. The items included the width of the shoulder strap for the camera unit, the placement of the control knobs for the camera unit and the positioning of the operating alert light on top of the camera unit.

Utility SF6 Management using MMW Con Edison has combined their Computerized Maintenance Management System, Maximo, with Maintenance Management Workstation (MMW) to track a SF6 usage. Masterwork orders have been created in the Maximo system for all pieces of SF6 equipment. When gas is added to a piece of equipment a sub work order is generated against the master. This provides the means to easily capture the number of times gas is added to the equipment. This information is then trended by MMW.

Temporary SF6 Leak Sealing To maximize the effective use of the information provided by the Laser Imaging Camera. Con Edison and Kinectrics (formerly Ontario Power Technologies) are working together to develop better non-intrusive methods to mitigate identified SF6 leaks. Through collaboration with EPRI, utilities are being approached to assist with field trails of the nonintrusive techniques. The effort will attempt to develop external sealing methods for the most typical areas of leakage in SF6 Circuit Breakers, such as bushing seals and "black pipe" assemblies of circuit breakers. Practical and effective external sealing methods of breaker assemblies are being developed for review and subsequent field trials. REFERENCES Boudene, C., et al, "Identification and Study of Some Properties of Compounds Resulting from Decomposition of SF6 under the Effect of Electric Arcing in Circuit Breakers", Rev. Generale de Electricite-Numero Special, June 1974. Chu, F. and Massey, R., "Thermal Decomposition of SF6 and SF6-Air Mixtures in Substation Environments", Gaseous Dielectrics III. Pergamon Press, 1982. Chu, F., “Procedures for Dealing with Arcing Byproducts from Gas-Insulated Substations”, Canadian Electrical Association Report, 128 T 239, April 1985. Dominelli, N. and Wylie, I., “SF6 Gas Condition Assessment and Decontamination”, EPRI Substation Equipment Diagnostics Conference IX, 2001. Dominelli, N. and Wylie, I., "Quantitative Field Detection of SF6 Decomposition Products", Minutes of the 60th Annual International Conference of Doble Clients, Sec. 10-7, 1993. Dominelli, N. and Wylie, I., “Analysis of SF6 Gas as a Diagnostic Technique for GIS”, EPRI Substation Equipment Diagnostics Conference IV, 1996. Dwyer, F.J., and Miners, K.B., "SF6 Gas Analysis - An Effective Means of Saving Time and Money", Presented at Southeastern Electric Exchange Annual Conference, April, 1982. EPRI Final Report. “SF6 Gas Condition Assessment and Decontamination”, 1000131. EPRI, Palo Alto, 2000. Mastroianni, M., "SF6 Analysis is the Key to Maintenance", Electrical World, Dec. 1980. Tominaga, S. et al, "Gas Analysis Technique and its Application for Evaluation of Internal Conditions in SF6 Equipment", IEEE PAS-100, Sept, 1981.

MMW - AIDING THE TASK OF EQUIPMENT CONDITION ASSESSMENT AT PEPCO Scott J. Buckley Potomac Electric Power Company (Pepco) 3400 Benning Road, NE Washington, D.C. 20019

ABSTRACT At the February 2001 EPRI Substation Equipment Diagnostics Conference, Tom Pierpoint (Pepco) and John Spare (KEMA Consulting) presented an overview of Pepco's Integrated Substation Equipment Condition Assessment Program.1 Part of that discussion summarized Pepco's utilization of EPRI's Maintenance Management Workstation (MMW) product. This paper is an expansion on this specific area, and illustrates several actual examples of the maintenance benefits that MMW and other computer-based tools are currently providing for Pepco’s Substation Operations and Maintenance Division.

INTRODUCTION As seems to be the case for just about every other utility in today’s marketplace, Pepco has been challenged to provide more reliability at less cost. As such, we need to obtain the best value possible for the expenditures of our precious maintenance dollars. To help meet this difficult objective, Pepco has re-baselined substation maintenance through an EPRI RCM initiative. From there, we are employing “living RCM” to continuously improve our substation maintenance program. Living RCM is providing us with a framework to change from a traditional time-based preventive maintenance (PM) strategy to one that is driven more by equipment condition.

STATUS OF PEPCO’S MAINTENANCE PRACTICES Since our presentation at this conference last year, Pepco has completed implementation of our core technologies and processes that are the “enablers” of our living RCM program. Pepco is now using the RCMdeveloped PM schedule as the trigger to perform a condition assessment of equipment before performing an internal PM. Depending on the findings of the condition assessment, the scheduled PM may be performed or may be rescheduled for a future date.

“Equipment Condition Assessment”, by Thomas Pierpoint and John Spare, EPRI IM&D Conference Proceedings, Feb. 2001 1

Some advantages of the condition-based maintenance model include: Reduces costs associated with performing unnecessary maintenance tasks Avoids forced equipment outages Leaves equipment untouched when invasive maintenance work is not needed, resulting in fewer maintenance-induced failures For some types of equipment we are using the same technology and processes to identify maintenance needs independently of the PM trigger. We will continue to expand this capability to more and more types of equipment. This will allow us to continue the transition into living RCM through condition-based maintenance.

THE ROLE OF MMW The MMW is an EPRI data-mining tool that is being developed as Pepco’s primary platform for its equipment condition assessment process and maintenance planning. It serves as a centralized source of information for equipment monitoring, and enables the maintenance department to trend failures, look for patterns in failure history, and monitor the overall performance of equipment.

OTHER EQUIPMENT CONDITION ASSESSMENT APPLICATIONS Two initial maintenance goals were targeted in Tom Pierpoint’s and John Spare’s paper: 1) a way to identify Load Tap Changers (LTCs) that have not passed through the neutral position, and 2) a tool that finds circuit breakers that have not operated in the recent past. Both of these tools have since been developed in house at Pepco and are currently being utilized by members of our Predictive Maintenance team.

OVERVIEW OF CONDITION ASSESSMENT TOOLS CURRENTLY BEING UTILIZED

MMW MMW includes the following features that make it a very powerful data-mining tool for performing equipment condition assessments: • •

Integrates all the key equipment condition data sources (Transformer Oil Analyst (TOA), EMS/SCADA Alarm File, X-Visor, PI Historian, Doble testing results) Retrieves equipment maintenance records and PM schedules from the Maximo Computerized Maintenance Management System (CMMS)

• •

Accesses status and analog points that are scanned by EMS/SCADA and archived in the PI Historian Makes this data readily available to all who are involved in equipment condition assessment and the associated maintenance planning functions

From an IT perspective, MMW is very empowering for end users since it serves as a common interface to multiple systems. Having the ability to use the MMW tool, a user is able to extract useful data from any system attached to MMW, even if they have never had any training on that specific system. Without a tool such as MMW, when a user needs to have a query or report built to get data out of a system, typically they are forced to seek the services of an IT specialist, and are therefore subjected to having to “wait in line” until the IT person gets a chance to work on their request. Pepco finds the MMW tool to be about as “point-and-click” as it gets in terms of building queries. It does not require users to be fluent in Structured Query Language (SQL). And the user interface is easier to use and more intuitive than is the case for most other systems, such as Maximo and PI Historian.

LOAD TAP CHANGER (LTC) OPERATION: NEUTRAL POSITION PASS CHECK If an LTC does not pass through neutral over a long time period, this can lead to particle buildup (a condition known as “coking”) on the reversing switch contacts. When an LTC passes through neutral, a mechanical wipe of the reversing switch contacts occurs, helping to keep them clean. This tool provides a means to identify any LTCs that have not passed through the neutral position, or haven’t moved at all, during a given time period. The Predictive Maintenance team uses this information to schedule the required maintenance task involving cycling the LTC through the range of tap positions. (See the “Detailed Designs of Internally Developed Tools” Section below for further details.)

LOAD TAP CHANGER (LTC) OPERATION: SLIDE-WIRE CHECK Pepco has experienced cases of faulty LTC position readings being reported by slide-wires. Since Pepco intends to increase its monitoring of LTC operations and utilize that data for the purpose of condition monitoring, the accuracy of slide-wire readings needs to be ensured. This database assists the maintenance team in locating potentially faulty slide-wires by showing a comparison of the number of tap position changes initiated by the EMS/SCADA system during a given time period, and the number of corresponding tap position changes reported in the PI Historian for the same time period. The EMS/SCADA alarm file does not show actual tap positions, but rather just indicates when a tap control request (raise/lower) has been issued to an LTC.

(See the “Detailed Designs of Internally Developed Tools” Section below for further details.)

BREAKER OPERATION CHECK Breakers that sit idle for an extended length of time (longer than 6 months to a year) are sometimes subject to either not tripping at all or tripping too slowly in the event they are called upon to operate. This is due primarily to the mechanism’s lubrication setting up and causing the equipment to malfunction. This database enables maintenance personnel to determine which circuit breakers have not operated during a given time period. Test trips are scheduled for breakers that are found to have had no trip events and therefore need to be exercised. This database also serves the purpose of showing the number of trips for each breaker that has had at least one trip event during a given time period. The number of trip events that a breaker has experienced is a key factor that is considered when prioritizing upcoming preventive maintenance work. The EMS/SCADA alarm file is the data source for this tool. This is because momentary trips do count as a breaker exercise event and, unlike PI Historian, momentary trips are captured in the EMS/SCADA alarm file. (See the “Detailed Designs of Internally Developed Tools” Section below for further details.)

DETAILED DESIGNS OF INTERNALLY DEVELOPED TOOLS

LOAD TAP CHANGER (LTC) OPERATION: NEUTRAL POSITION PASS CHECK This is a PI Historian query. All of the LTCs that are scanned by EMS/SCADA and archived in the PI Historian are listed down the left-hand column (“TTP” is the standard naming convention for Pepco’s tap position points). To run the query, the user only needs to edit the date range that is desired to be covered. The Minimum and Maximum functions within PI Historian are utilized to return the min and max tap position values for each LTC during the given time period. If the range of tap positions for a particular LTC does not span the neutral position, then “No” is written into the “Neutral Pass?” column and the cell is colored red so that it can be easily spotted as the user scrolls through the spreadsheet. The user can simply choose to sort or filter by the “Neutral Pass?” field if they wish to see all of the problematic LTCs grouped separately from those LTCs that have passed through neutral. The “Position Change?” field operates in much the same way, except it uses the color blue instead of red to highlight the cells.

LOAD TAP CHANGER (LTC) OPERATION: SLIDE-WIRE CHECK This tool is actually composed of two separate queries that extract data from two separate sources – one query provides LTC data from the EMS/SCADA alarm file, and the other provides LTC data from the PI Historian. The “SCAD Tap Move Count” data is provided by a separate MS Access database that is tied to the EMS/SCADA alarm file, and this MS Access database is linked to the spreadsheet via a MS Query. The “PI Tap Position Count” data is provided by a PI Historian query. Similar to the above LTC query, the entire list of LTC point names is hard-coded into the query. When a user runs this query by changing the date range, two things happen. First, the entered date range gets fed into the underlying MS Access query which creates a list of all tap changer operations that it finds in the EMS/SCADA alarm file for that time period, then the MS query groups that list by RTU and point name, and calculates the number of LTC control requests issued for each LTC. Second, the entered date range is fed into the PI Historian query to calculate the number of tap position changes that occurred for each LTC during that time period.

LOAD TAP CHANGER (LTC) OPERATION: SLIDE-WIRE CHECK

BREAKER OPERATION CHECK The breaker operations tool is a MS Access database linked to the EMS/SCADA alarm file. Also, it includes a table imported from a PI Historian query that lists all of the breaker names that exist in the system. In order to determine what breakers have not operated during a specific time period, the query works by first determining the entire list of breaker names that have at least one trip operation showing in the EMS/SCADA alarm file during the given time period. It then uses this list in a comparison against the list of all breakers existing in the PI Historian system, and in this way is able to produce a list of breaker names that have not been operated. Below is the initial screen that the user receives when they launch the application:

If the user chooses to view the trip count list, the following screen appears:

The user enters the date range that they are interested in covering, and then clicks “Run Query” and the results appear as below: (Note: You will notice that there is a field on the above screen showing the most recent date available in table SCAD. This exists to inform the user as to what data is currently available for querying. The EMS/SCADA alarm file that we have access to is not real time – there can be some lag between when an event occurs in EMS and when that event history gets copied into the EMS/SCADA alarm file.)

The “Total Trips” field indicates how many times a particular breaker opened. The “CN_Initiated” field shows how many of those trips were the result of an EMS console-initiated command. The “Momentaries” field indicates how many of the trips were momentary operations.

If the user instead chooses to run the query that lists the breakers that have not tripped, the screens are as follows:

The user enters the date that they want the query to start from, and clicks “Run Query” and the following list of breakers without trips appears:

OCB Diagnostics Rick Youngblood, Cinergy Fredi Jakob, Karl Jakob & Simon Jones, Weidmann-ACTI Alex Salinas, Southern California Edison I. Introduction During the B.D. era, before deregulation, run to failure, time based and operation count based maintenance methods were widely employed. These methods were effective in maintaining the power delivery system but were labor intensive and not cost effective. Time based and operation count based maintenance methods could not identify units that developed problems between scheduled inspections. Units were often inspected on a time basis and no problems were identified. After deregulation, A.D., fiscal requirements led to decreases in trained maintenance personnel, deferral of capital expenditures, equipment life extension programs, and efforts to maximize uptime and minimize maintenance costs. Reliability centered and condition based maintenance are key A.D. concepts that have been implemented in the power industry. The ultimate goal of condition based maintenance would be to perform maintenance “just in time”, before the equipment fails in service. Condition based maintenance requires periodic or continuous (on-line) equipment monitoring. Equipment is scheduled for inspection and/or maintenance only when diagnostic test results indicate a potential problem. II. OCB Maintenance Oil filled circuit breakers; OCB’s have traditionally been serviced on a time based, operational count or fault current basis. We believe that OCB condition can be ascertained by non-invasive diagnostic tests that can reliably indicate when an internal inspection and/or maintenance or repair is required. Cinergy, Southern California Edison and Weidmann-ACTI have cooperated in a project to identify a battery of chemical and physical tests to assess OCB condition. III. Diagnostics Development of our diagnostic program involved four sequential steps: 1) Identification of fault mechanisms 2) Selection of appropriate physical and chemical tests to detect problem units 3) Correlation of test data with physical inspection of problem units and 4) Development of algorithms to classify OCB condition. A. Fault Types OCB contacts can become coated with oxides or sulfides, which result in increased contact resistance and increased contact operating temperatures. Contacts also erode because of mechanical wear or arcing. Arc suppression grids, which are made of cellulose, deteriorate to some degree whenever an arc is suppressed. Each of these OCB problems can be detected with one or more diagnostic tests. Figure 1 shows contacts that have been severely eroded. Figure 2 illustrates a new arc suppression grid and one that shows degradation. A wide range of chemical and physical test methods is available to detect these types of faults.

Eroded Moveable Contacts Figure 1

Close up of a Degraded Grid.

“Good” Grid

Figure 2

B. Laboratory Tests

Dissolved Gas Analysis, DGA, has been extensively applied to locate incipient thermal or electrical faults in transformers. Normal OCB operation will produce the “key gases” associated with arcing under oil. These key gases include acetylene and hydrogen, which are produced at the very high temperatures associated with arcing. The “hot metal gases”, methane, ethane and ethylene are produced whenever the oil is overheated from any cause.

The temperature required to produce acetylene is considerably higher than that required to produce the hot metal gases. Insulating fluids absorb and distribute fault energy. Thus the temperature of the oil is very high in the vicinity of an arc and decreases with distance away from the arc. This variation of temperature results in the production of acetylene close to the arc and hot metal gases further removed from the arc location. Very little heating occurs in a healthy OCB so the amount of hot metal gases generated should be small. This analysis of key gas production mechanisms indicates that the ratio of heating to arcing gases should be indicative of problems such as increased contact resistance or contact erosion. Arc suppression grids deteriorate every time that they quench an arc. When the grids are new the arc suppression time is low. The particles generated from the grid degradation are numerous but small in size. As the grid opening enlarges the arc suppression time increases and larger sized degradation particles are produced. A fault is limited only by the maximum current amplitude of the source impedance and the interruption time of the breaker. Maximum fault power (I2T) determines time to failure. Particle size and count continues to grow until the distance from the point source of heat and the maximum grid hole diameter cause a cooling and blast zone buffer. At this point the production of larger particles decrease and smaller particles again increase. A plot of large particle size production with time seems to follow a bell shaped curve. The difference in the leading edge to the trailing edge of the curve is noted by the measurable increase in contact metals present in the oil. Arc suppression grids are constructed of cellulose and these particles can be distinguished from carbon and shiny metal particles by chemical microscopy. At temperatures above 300o C cellulose is destroyed and the resulting carbon particles are observed in the oil. Arc tip and arc shaft erosion can be measured by determination of the characteristic metals in oil and with chemical microscopy, which can distinguish between shiny metal and carbon particles. Oil quality assessment tests are also useful in identifying OCB problems. Dielectric breakdown voltage measurements, ASTM D-1816, are effected by moisture, metals, carbon particles and cellulose particles in the oil. Particles, especially carbon, also effect oil color. Since OCB’s are free breathing devices the moisture level is higher than that found in sealed components. Table 1 is a summary of the tests that we have incorporated in our OCB diagnostic program and the type of problem that can be determined with these technologies.

Table 1 Diagnostic Tests OCB Problem Increased Contact Resistance

Test(s) DGA

Contact Tip Erosion

DGA, Metals, Chemical Microscopy, Dielectric-1816

Arc Suppression Grid Degradation

DGA, Particle Count, Chemical Microscopy, Dielectric-1816, Color

Result(s) Increased Hot Metal Gases, Increased Heating to Arcing Gas Ratios Hot Metal Gases, Metals Observed in Oil, Metal Particles Observed by Chemical Microscopy, Lowered Dielectric Breakdown Voltage Increased Fault Gas Levels, Large Particles in the Oil, Cellulose Fibers, Decreased Dielectric Breakdown Voltage

IV. Diagnostic Protocols Once the appropriate analytical procedures were selected to ascertain the effect of OCB problems it became necessary to determine normal values for each of the measured parameters. We used a statistical approach to evaluate test data from problem free units in order to establish the norms. We then determined the ninetieth percentile values for each parameter, for several thousand OCB samples. For example, the 90th percentile individual fault gas concentration and the total dissolved combustible gas concentration for all samples evaluated are given in Table 2. Data for selected heating to arcing fault gas ratios was calculated in a similar manner. Similar calculations combined with field observations were used to establish norms for particle size distributions and metals in oil. Relevant IEEE guides are used to evaluate results of oil condition assessment tests. These values are all generic in nature and no attempt has made at this time to develop unit specific flag points. Table 2 90th Percentile Gas Concentrations in OCB Oil Samples Fault Gas

90th Percentile Concentration

Hydrogen Carbon Monoxide

62 136

Methane Ethane

28 14

Ethylene Acetylene Total Dissolved Combustible Gas

71 173 530

Diagnostic software has been developed to evaluate OCB samples according to our established norm values. OCB sample results are placed into three broad categories of Normal, Caution and Warning. Southern California Edison designates OCB condition numerically using a 3, 2, and 1 scale. Normal indicates that the sample should be retested according to the initial utility criteria. Caution indicates that the sample should be tested more frequently, again at a rate determined by utility protocols. Warning indicates that an internal inspection is appropriate. Tables 3, 4 and 5 are examples of test reports for samples that fit in each of these evaluation categories.

Table 3 OCB Condition: Normal (3) Serial number Model Pole Manufacturer Date

Hydrogen Methane Ethane Ethylene Acetylene Carbon Monoxide Carbon Dioxide Oxygen Nitrogen Total Dissolved Gas Total Combustible Gas Equivalent TCG %

20695 TK1 C FPE

Counter # Gallons Fluid Type KV Rating Interrupt

267 Mineral 69

Condition Code: Normal (3)

2 1 0 1 0 84

Between 2 & 5: Between 5 & 15: Between 15 & 25: Between 25 & 50: Between 50 & 100: Greater than 100:

3374895 749850 4084 571 90 0

Moisture Color Number Dielectric 1816 Silver Chromium Copper

28 L2.0 20 1-24 L0.50 L0.50 L0.50

518 24309 53664 78579

ISO Code: Fibers: Metals: Carbon:

23/20/13 20 10 50

Nickel Phosphorous Lead Tin

L0.50 L0.50 L0.50 L0.50

88

Other

20

Zinc

L0.50

0.0973

Opacity

2.0

Tungsten

L0.50

Narrative: No problems found. No action taken. Unit stayed in service, reset maintenance schedule.

Table 4 OCB Condition: Caution (2) Serial number Model Pole Manufacturer Date

28760 CG-38 MGE 1987

Counter # Gallons Fluid Type KV Rating Interrupt

120 220 Mineral 38 40

Hydrogen Methane

1 1

Between 2 & 5: Between 5 & 15:

2199309 3891922

Ethane

1

Between 15 & 25:

98949

Ethylene Acetylene Carbon Monoxide Carbon Dioxide Oxygen Nitrogen Total Dissolved Gas Total Combustible Gas Equivalent TCG %

2 2 17

Between 25 & 50: Between 50 & 100: Greater than 100:

382 25182 50557 76145

Condition Code: Caution (2)

30 L3.0

2252 60 0

Moisture Color Number Dielectric 1816 Silver Chromium Copper

ISO Code: Fibers: Metals: Carbon:

23/22/17 10 5 70

Nickel Phosphorous Lead Tin

L0.50 L0.50 L0.50 L0.50

24

Other

15

Zinc

L0.50

0.0229

Opacity

4.0

Tungsten

L0.50

Narrative: Analysis indicated mild contact erosion, and poor oil quality. Remedial Action: Unit was put on ½ maintenance schedule and remained in service.

10 1-24 2.38 L0.50 L0.50

Table 5 OCB Condition: Warning (1) Serial number Model Pole Manufacturer Date

Hydrogen Methane Ethane Ethylene Acetylene Carbon Monoxide Carbon Dioxide Oxygen Nitrogen Total Dissolved Gas Total Combustible Gas Equivalent TCG %

318027-A FZO-69 A Allis1953

Counter # Gallons Fluid Type KV Rating Interrupt

608 855 Mineral 57 21

Condition Code: Warning (1)

7280 6624 729 5424 9356 91

Between 2 & 5: Between 5 & 15: Between 15 & 25: Between 25 & 50: Between 50 & 100: Greater than 100:

901411 3453273 2645616 221381 2462 0

Moisture Color Number Dielectric 1816 Silver Chromium Copper

23 L6.5 10 1-24 L0.50 L0.50 1.63

437 44520 24023 94686

ISO Code: Fibers: Metals: Carbon:

29/28/24 10 10 80

Nickel Phosphorous Lead Tin

L0.50 L0.50 L0.50 L0.50

25706

Other

0

Zinc

L0.50

20.6182

Opacity

5.0

Tungsten

L0.50

Narrative: Oil sample obtained approximately 4 hours after a 30000 AMP, 3 cycle fault. Contacts evaluated at 25 percent degraded. Copper present due to shaft wear on movable contact. Grid not significantly degraded. Remedial Action: Rotated movable contact ¼ turn. Filtered and processed oil. General maintenance. Returned to service. V.

Sampling Protocols

The overall objective of our OCB diagnostic program is to minimize the number of required internal inspections and thus to realize considerable maintenance cost savings. Utilities must decide when and at what frequency to sample their OCB’s. A sampling strategy has been developed by Mr. Alex Salinas at Southern California Edison, SCE, to determine when to draw the initial sample, when to draw subsequent samples and when to conduct an internal inspection. The entire sampling process is incorporated in their system software. The software, which maintains OCB operations data,

generates an initial work order when any of three conditions are met: twenty interruptions, three hundred operations or five years from the last test. This first work order will be for an initial oil sample. The results of the initial test generate a code: 1=Schedule internal inspection, 2=Reset oil sample trigger at half the initial values (10 interruptions or 150 operations) and 3=Reset sample trigger at 20 interruptions or 300 operations. The software maintains all of the test data, resets the triggers and prints the work orders. This closed loop operation requires no operator intervention. A flow chart for the SCE protocol is given in Figure 3.

Figure 3 SCE Flow Chart Maintenance cost savings for the SCE program can be estimated considering the number of reduced internal inspections. SCE currently inspects about three hundred breakers a year, both oil filled and others. The oil filled breakers range in voltage class from 7.2 kV to 220 kV. Their largest number of breakers operates at 69 kV. Inspection costs range from 2K$ - 20K$ on OCB’s in the voltage range from 7.2 kV to 220 kV. Based on statistical evidence less than ten percent of the units tested will require an internal inspection. The savings promise to be considerable. Cinergy Corporation currently uses both condition based (CB) and fault adjusted operation count (FAO) as the initial triggers for breaker maintenance. The non-invasive condition based triggers used are

thermography, ultrasonics, oil quality and dissolved gas analysis. The main invasive tests are ductor and power factor. The invasive tests are used minimally due to limited outage request acceptance. Fault adjusted operation count is a method that employs knowledge of the maximum fault amplitude based on source and circuit impedance and fault duration. This is used to determine the estimated number of breaker operations before maintenance is needed. Unless the breaker employs the use of some form of I2T monitors it is still very difficult to determine true fault duty cycle. Additionally, due to deregulation and customer commitment, outages and internal inspections will be kept to a minimum. Although each test gives different positive information to determine maintenance interval, Cinergy has found the addition of oil particle analysis to determine grid and contact health invaluable to determine overall maintenance interval. VI. Conclusions and Future Work. Normal or threshold values for OCB test parameters are at this time generic in nature. Because of design variations unit specific normal values may be more appropriate. Unit specific test values will have to be empirically determined. We are maintaining a very detailed database so that we can develop unit specific normal values in the future. We are cooperating with utility clients to further evaluate testing frequency protocols. For example SCE tests the OCB oil every five years if the interruption (20) or operation (300) criteria are not met. Can this time interval be extended without increasing system outages due to breaker failures?

LIFE MANAGEMENT OF STATION BATTERIES THROUGH CELL MANAGEMENT Gerald W. (Gary) McDermott U.S. Bureau of Reclamation, Curecanti Field Division, Montrose, Colorado

Over the years the batteries that are used in power plants and substations have not changed a lot. Yes, we have batteries made of different material, some of them are sealed, and others use different electrolytes. The battery system usually forms the heart of the DC control and protection system for the station. This DC system is the power source for equipment control and protection. The need to have this system extremely reliable is what drives preventative maintenance on batteries. Regardless of the type of battery, it must be maintained to function properly. We typically maintain our batteries on a weekly, monthly, and yearly schedule. On a weekly cycle, the battery is inspected, several cell voltages are read, and the respective specific gravities are checked. On a monthly basis, all the cell voltages are read and all the specific gravity readings are taken. Annually, all the inter cell connections are removed, cleaned, lubricated and reinstalled. Doing all of the recommended maintenance requires many man-hours. At the US Bureau of Reclamation Morrow Point hydroelectric plant, we spent about 8 hours per month doing maintenance on the batteries, and a full day doing the yearly maintenance. This amounts to about 104 man-hours, or 13 man-days a year or about $5,200 per year. Compared to other equipment in the plant, this makes the battery a relative inexpensive maintenance item; however, it is a significant cost when compared to the cost of a battery. After 20 years and $104,000 worth of maintenance, we could have replaced the battery several times. After we do all the maintenance on the battery do we really know how well the battery will perform under actual operating conditions? As we found out at Morrow Point, we may have actually been creating problems. At the Morrow Point power plant a new battery was installed four years ago. The battery is a 58 cell, 420 Amp-hour unit that was operated on float-only charge at 130.5 VDC. The battery was installed according to the manufacturer’s instructions and put in service. About two months later we installed a battery condition monitoring system. Reclamation’s research group recommended the system we purchased. It was new on the market and had been developed by MCM Enterprise Ltd. through an EPRI research project. The system was chosen because of the features it has and its ability to give long-term trends on the individual battery parameters. We felt that any system we installed had to have the following minimum features: 1. Be non-intrusive to the battery case. 2. Software and hardware compatible with Windows NT operating system. 3. Be able to provide the following readings: a. Cell voltage. b. Cell Specific Gravity. c. Test the integrity of the cell-to-cell connections. d. Monitor the amount of cell float current necessary to maintain a uniform charge on each cells e. Give the direction and magnitude of the bank current. f. Bank voltage. g. Automatic monitoring of the bank during discharge.

The installation of the monitoring system consisted of installing 2 sensing modules to the side of each battery jar. These are attached with a self-adhering tape that is an integral part of each cell module. In addition, it was decided to install a wire gutter to

contain the cell-to-cell data link and other wire. This made for a neat installation. See Figure 1.

chokes Measuring leads

Figure 1 shows the sensing modules and wire gutter attached to the jars. Also shown are the measuring leads attached to the bus bars, and the chokes attached to the main battery leads.

Bank monitor/system controller

Communications interface

Current sensor

Figure 2 shows the communication interface (with built in modem) and the bank monitor/controller mounted on the wall, the current sensor mounted at cell #1, and the rest of the hardware. Next, connections were installed to each bus between the battery posts. These connections are used to measure cell voltage, and voltage drop across the cell when a one-millisecond bank load is applied for the current path integrity test. A choke (isolation inductor) is installed around each main battery lead. The installation of all the equipment took two days and consisted of mounting modules on each cell, installing power cabling, and mounting the bank monitor and communications interface equipment. See Figures 2.

The overall system is shown schematically in Figure 3. The data links between the cell monitoring modules are optically isolated to eliminate the possibility for a current path to exist between modules. The isolation inductors (chokes) provide test isolation for approximately one millisecond during the connection resistance test (current path integrity test).

Figure 3 shows the overall configuration of the non-invasive battery monitoring system

Once the equipment was installed the system was set up and calibrated. A complete set of readings was taken on the battery and the data entered into the software as the starting point for monitoring. Much to our surprise, within days of putting it in service, the monitoring system started to report problems with the battery. The first deficiency it found was high resistance at several inter-cell link and battery post connections. We had taken resistance readings on the connections and not found any problems. However, the monitor takes connection voltage-drop readings while a load of about 400 amps is applied to the battery for one millisecond (current path integrity test). This method found poor connections that we had missed. See Figure 4.

Figure 4 is a view of the opening page showing cells with problem in red. All the cells in question had high resistance connections.

The current path integrity is measured by employing the four-wire resistance technique. To prepare for the measurement, the bank monitor simultaneously instructs all cell monitors to measure their cell’s voltage and store the result in memory. The bank monitor next switches an electronic load across the bank that draws current for approximately one millisecond. At the peak of the bank current, which is arranged to be roughly equal to the amp-hour capacity of the bank, the bank monitor again simultaneously instructs all cell monitors to re-measure cell voltage and subtract the results from the zero-current values of voltage already stored in memory. The difference voltage for each cell, reported as “delta V” is proportional to the

Voltage-current relationship for measuring the voltage drop across connection resistance in the series circuit

The load is electronically inserted across the battery for one millisecond. The chokes provide test isolation.

Figure 5 shows the process for automatically measuring current path integrity

resistance of the current path including both the cell and strap/post connections. Figure 5 shows the relationship of the voltages and currents during this test. Upon opening the software up to the individual cells, one can get either a cell-bycell view of the battery or a view that shows the average cell, the high cell, and the low cell in each category. In looking at Figure 6 you can see the trends of cell #1 over a period of time. Of interest is the gradual rise in the voltage drop across the connection of this cell. With every voltage drop alarm, our electricians reported that

the connecting bolts did tighten before the torque wrench reached the desired value. It was also interesting to note that just prior to developing a high resistance connection the battery underwent a load cycle as indicated by the downward spike in the readings, shown in Figure 6.

Figure 6 gives a view of the screen that gives trends on individual cells.

After some months of operation MCM reviewed the data that was being taken. They noticed that the water consumption of the battery was fairly high. From the Float Current Bypass cell readings it was noted that all the cells were bypassing 100% of the float current. In essence this indicated that the battery was being over charged, making it use more water. We started lowering the float voltage until most of the cells read in the 80% bypass range. This reduced water consumption in the battery. See Figure 7. The system allows the user to look at many conditions in each cell, and the battery bank as a whole. In addition it will record bank discharges, showing the overall bank and each cell’s performance during the discharges. This information goes a long way to assuring the quality and integrity of the battery. The monitoring system performs a test on the battery however often the user desires. The monitoring system is set up to automatically do a test and collect data every 24 hours at Morrow Point. This gives us good data without building a huge database. However, during a load test on the battery or during an actual emergency event, the battery monitoring system steps up its data measurement rate to capture the performance of the battery, during the event. The data is downloaded to a computer in the plant once a week, reviewed, and any necessary actions taken. This test sequence yields the data that we have looked at above. This system has been in use for almost two years, and has successfully found serious problems; it has also minimized the required battery maintenance. The cost of this system approaches the cost of a battery. In order to justify this cost we must take a close look at what our battery does, and how important it is in the overall scheme of things. Not all battery systems are used for the same purpose, however most of them operate the control and protection schemes for large rotating equipment and breakers when they are employed in power plants and substations.

Figure #7 is a screen view showing the high, low and average of all the cells. Note that the fluid levels move down. The straight line shows where water was added. Note place where the bank voltage is lowered you also notice that the slope of the water use curve changes, and the cell bypass currents drop.

For the most part the DC control and protection circuits in Reclamation facilities run on battery chargers most of the time. Only under charger failure or loss of station AC power do we need the batteries. As often happens, it is during times of system or equipment problems when station AC disturbances will occur. This is simply Murphy’s Law in action. These are the times when controls, protection and breakers MUST operate to prevent damage to equipment. Thus, the main purpose of the battery is to provide critical power to control and protection equipment when it most needs it. Failure of the battery system during these periods of time can have consequences that range into the millions of dollars for the equipment they protect. Thus it must be argued that the justification for the monitoring system should be based on the cost of these consequences rather than the ratio of the cost of the battery versus the monitoring system. The bottom line for the monitoring system at Morrow Point is that it has drastically cut our man-hours of maintenance. Since the monitor’s installation, several bad connections have been corrected, and the battery bank has had water added only twice. Then since system has been in service; a set of readings was manually taken after one year to verify the readings of the monitoring system, and they proved to be identical. The Morrow Point crew has decided that a lot of the maintenance that was done previously was probably causing as many problems as they solved. Redoing the jumper bar connections probably loosened more connections than it did to repair possible corrosion and tightness problems. The need to be assured that your battery will operate as designed when it is needed most has been assured. This along with the reduced maintenance makes the cost of the system easily justifiable.

IMPLEMENTING SUBSTATION PREDICTIVE MAINTENANCE

Jon L. Giesecke, Project Manager EPRIsolutions. Lenox Engineering and Test Center 115 East New Lenox Road Lenox, MA 01240 [email protected]

EPRIsolutions’ Substation Predictive Maintenance Program (SPdM) was developed to satisfy the need to know equipment condition. Aging equipment and reduction of the work force, makes this program an excellent choice for a utility that wants to “keep the lights on” now and into the future. The SPdM program strategy is to review a utility’s existing maintenance program and then to compliment their maintenance efforts by simultaneously applying a series of condition monitoring tests. Every utility has a different set of needs. The SPdM program is customized to fit each utility. Transformers, circuit breakers and support equipment such as current transformers and potential transformers are monitored. The following elements are used during the development of the SPdM program. • • • • • • • • • • •

Assessment Process Field and classroom training in all aspects of the program Infrared Thermography Inspection Ultrasonic inspection for partial discharge and other internal transformer problems Airborne ultrasonic inspection Sound & Vibration spectrum analysis of energized transformers Review of DGA and oil quality results Circuit breaker condition monitoring Day time corona detection SF6 leak detection Communication and Data Management

Program Goal – Eliminate all Failures SPdM Survey Scope The SPdM Survey encompasses all substation / switchyard equipment considered vital to the operation of the plant regardless of ownership. Survey results are communicated through a Management Summary Report, which includes: •

Critical Areas of Concern -Lists items which warrant immediate attention.

•

Equipment / Technology Matrix -What technology is applied to each type of component.

•

Condition Status Report -Provides list of equipment issues and rates severity of problems found.

•

Avoided Cost Summary -Conservative dollar value of the avoided cost if the problem had not been identified during the SPdM survey.

Scope of Inspections and Testing performed during the Survey includes: Perform a visual inspection of switchyard components -Looks for oil leaks, cracked / dirty insulators and bushings, heat discoloration of equipment, proper oil levels, and pump flow indicators. Listen for unusual sounds and vibration. Functional testing of radiator fans and oil pumps -ensures proper airflow and direction. Pumps are inspected with vibration, thermography and ultrasonics. Ensures control circuit to fans and pumps will work when needed Thermography inspections -Ensures no hot spots exist on transformers, breakers, disconnects, bolted connections and control circuits. Verifies minimum temperature profile across transformer and LTC. Provides verification of oil winding temperature gauge. Verifies minimum temperature differential between Transformer and LTC. Verifies no blockages and proper operation of radiators. Sonic / Ultrasonic inspection of transformers -detects partial discharge and arcing inside transformers and LTC’s. Readings are evaluated against DGA results and recorded for trending. Vibration inspection of transformers -detects changing conditions (core looseness, broken clamping, or deformed windings) inside transformers. Readings are evaluated and recorded for trending. Airborne Ultrasonics and sound levels -detects ionization of air around connections, bushings and insulators. Detects background noise levels around critical components such as transformers. Readings are evaluated and recorded for trending. Issues found are resurveyed with complementary technologies

including thermography, vibration, and contact ultrasonics where possible. Follow-up can be performed with Corona camera. Review of Oil quality and DGA readings -Evaluates for trends and used to verify online DGA equipment readings where installed.

Program Management Prior to beginning or expanding an SPdM program, you must determine where you are. A properly executed predictive maintenance process is not just the practice of using diagnostics to detect equipment faults or degradation. The program should be a direct extension of means to meet goals set forth by management - reducing costs and improving performance through: • Prioritizing and redirecting resources to equipment in need. • Reducing the scope and corresponding costs of preventive and corrective maintenance tasks. • Reducing the amount and/or frequency of corrective and preventive maintenance tasks. • Extending equipment life. • Increasing availability and performance of equipment. Experience shows that it could take several years to establish a robust SPdM process. Program development and implementation requires an investment in new test equipment, personnel training, changes to roles, goals and responsibilities of individuals and organizations, new lines of communication, development of procedures, and support of changes to the cultural values of the organization. Initiating or enhancing an SPdM program is a six-step process, using the following steps: 1. Determine exactly how SPdM will support the achievement of financial, safety, reliability, and regulatory or other goals. 2. Establish the program formally. Place an individual in charge and develop specific goals for the SPdM programmer in the areas of organizational readiness, identification of components and application of technologies. 3. Formulate a detailed Action Plan with task lists and a schedule for completion. 4. Execute the plan by procuring equipment, educating personnel in roles and responsibilities and technology applications. 5. Create measures and means to track and evaluate program effectiveness. 6. Automate the process through the use of analysis software and electronic data entry.

Assessment Process Experience with implementing the SPdM process showed that many organizations used existing technologies effectively, but were lacking in the areas of program coordination. Alternately, some had programs well coordinated but lacked the effective use of technologies. A listing of assessment areas considered to be “best practice” observations can be used to self-assess the implementation process or grade the progress against industry standards. The assessment looks at the topical “best practice” areas and at the same time, takes a “systems” view of the organization - looking at how technologies, people and processes were integrated and optimized to achieve the goals of the program.

Topical areas include: Equipment Condition Indicator Matrix

Training Program

Technology Applications

Schedules

SPdM Process/Work Flow Diagrams

Quantifiable Goals

PdM Supervisor

Financial Basis

Organization Roles and Responsibility Charts

Metrics/Cost Benefit Tracking

Communication Formats

Customer Satisfaction

Equipment Condition Status Reports

Continuous Improvement

Each of the above elements has question sets, which are used during the assessment process. In turn this allows the assessment team to assign a numeric score. An assessment identifies strengths and weaknesses of a program and details an action plan for improving the program. Figure 1 is a ‘Spider Chart’, which is a graphic representation of example assessment results. An average score of 80% is considered world class. Initially, and one year after formally adopting the condition based maintenance philosophy, are good times to assess program progress.

Switchyard/Substation SPdM Program Assessment Spider Chart Evaluation

Equipment Condition Indicator Matrix 10.00 Technology Applications 8.00

Continuous Improvement

Customer Satisfaction

WORLD CLASS

6.00

PDM Process/Work Flow Diagrams

4.00

Metrics/Cost Benefit Tracking

2.00

PDM supervisor

0.00 Organization\Roles and Responsibility Charts

Financial Basis/Plan

Quantifiable Goals

Communication Formats

Schedules

Equipment Condition Status Report Training Program

Figure 1- Graphic Representation of Example Program Assessment Results

Avoided costs The SPdM program is now in its tenth year. It began as a two-year research project with one utility. Many utilities have successfully participated in this program. The equipment saves and avoided costs, through condition monitoring and trending, have been abundant. The combined cost benefits are rapidly approaching the $50M mark. Estimates of the avoided cost benefits for utilities implementing SPdM have ranged from 12% to 200% of total annual maintenance budgets. These numbers vary widely due to the nature of the utility size and the significance of the findings (i.e. a utility with an annual substation maintenance budget of $500,000 finds a transformer in distress – estimated cost benefit may exceed $1M). They also vary due to the differences in dollars allocated for substation maintenance ($0.15 - $0.75 per MWh of energy processed through the substation). Most commonly, utilities report that the program has paid for itself during the first supervised training session.

4 COMMUNICATIONS/DATA MANAGEMENT/SYSTEM INTEGRATION CHAIR Chair: Michael Ingram, Tennessee Valley Authority

Data Standardization Tony McGrail The National Grid Company, plc [email protected]

Abstract As data sources and applications proliferate, what can maintenance engineers do to keep track of important data, and ensure the smooth running of vital software? Work currently ongoing at EPRI looks at data standardization for substation related data through extension of the Common Information Model (CIM), and application of an integration bus; simple guidelines for design and development of substation related software are being applied. Using this approach, data would be easily communicated between applications, and made available to any authorized user or application. The specification, production, installation and management of software becomes simpler and cheaper with consequent reduction in IT spend.

Background Over time, software products have been developed to address particular needs. Making best use of available technologies useful tools for engineers in the power industry have been developed. However, software products have been developed independently, mirroring the approach taken in other industries, and in accordance with prevailing best practice in software development. This approach leads to software working independently and with different data formats. Further, distinct approaches required in different developments have lead to inconsistencies in presentation, user interaction and product support. Consequently it has been difficult to implement software packages in a coordinated and seamless manner. This is a common problem for software developers and for many industries, but techniques are available which address this issue – moving from individual software packages to a more holistic approach. This requires separation out of data from algorithms that work on the data, combined with a consistent approach to presentation – the so-called three-layer architecture described in this report. This paper outlines a strategy that will allow current software to move from its present individualistic approach to a more consistent and seamless environment where implementation is easier, and software becomes plug and play.

Vision The vision is one where software: •

is easy to specify and quick to prototype

•

is simple to install and use with a minimum of administration involvement

•

implementation does not require duplicating presently held data

The vision is one where data: •

is stored and communicated in standard formats

•

is accessed through standard communications protocols

•

is published and made available for any user or software application

A means of achieving these visions is available, however, a look at software, data and related issues will be useful in providing a context.

Software Development Software architectures have developed over time to address specific needs, and in response to specific drivers. The move from batch to interactive and GUI programming has resulted in a proliferation of databases and applications.

Simple Software – Single Application/Single User Early software combined and intertwined data and algorithms in a way that was most efficient for calculations on low memory computers. Changing values in a calculation required re-entry of the software into the computer. At this point, results of a software program would be either stored to file or presented in a rudimentary way on a user screen.

Data Separation Separating out a portion of the program to cater for data meant that the two could be managed independently. Much work in this area was pursued by Codd, in the late 1960’s at IBM; this was subsequently taken up by Date [1], and the notion of relational databases was born. This advance meant that a given application in software could be applied to a number of different datasets without re-entry of software into the computer. This gave a great deal more flexibility in software development and use, and data was beginning to be seen as a separate entity.

Data Formats Note the separation out of data does not necessarily mean that the data will not provide problems in its own right. Indeed data separation does not solve any problems as regards redundancy or inconsistency – the two key problems with data as noted by Codd in the late 1960’s. The growth of formal database design, in particular the analysis of source data in terms of entities and relationships has lead to a formalization of database design that is both practical and efficient.

Growth of GUIs and User Interfaces Software architecture developed slowly, as languages proliferated and structured programming for batch mode program came to the fore. With the development of graphical user interfaces (GUIs) and personal computing, software became available for individuals on desktops. At this time the proliferation of databases begins to be a problem, and the concept of software version control and access to data began to grow uncontrollably.

One positive development was a consistent approach to presentation under, for example, a Windows environment. Even here, this does not predefine the way in which a software package will present results, but it does provide a degree of formalism. The dashed line around the application and presentation layers shown in the figure below, represent the fact that the two layers are still intimately related. This approach does start to address issues of software control, but these are only partial solutions. The situation is exacerbated when there are multiple software packages on a network accessing multiple databases, possibly communicating data between applications: Data is now distributed in different databases – possibly with duplication and inconsistency. Dealing with data and multiple software versions has been partially solved through application of a three-layer architecture. At this point software may still have data embedded into it, and is built for individual purposes, deployed on local PCs. Client server applications began to address some of the issues involved in distributed data bases and multiple versions of software.

Simple Three-Layer Architecture The three-layer architecture builds on previous examples, but formally separates out the three layers. Applications may now be written as library modules, with the algorithm embedded, even encrypted, into a small executable file. The presentation (gui) and data layers are formally separated. Data is still stored in a number of databases, but these are distinct from the software algorithms and applications that act on the data. Where an algorithm produces a report, this may be presented on screen, published as a web page, or stored as further data for future analysis. Clearly, this architecture does not preclude the possibility of data duplication. What it does do is allow the algorithms and applications that add value to access appropriate data, making it easier to switch between software packages. It is akin to a choice between Word and Word Perfect, for example. By taking this analogy one step further, data standardization would mean that there would be no conversion required between Word and Word Perfect documents – they would both be stored in the same format. The benefit of one package over another is its application, not in the way it stores the data.

This preceded developments of corporate intranets, which took client server applications to the next stage of development. Instead of a client-server application, a single application was made available over the intranet, without any application specific software required at the user PC. By accessing well-managed data resources, through intranet enabled software, a consistent and simple approach to software delivery can be achieved.

Data If a utility doesn’t have its data ducks lined up, there will be inevitable problems in extracting value from the data. Good applications and infrastructure can help paper over the cracks, but their needs to be good data management and maintenance for the software to thrive and survive. GIGO holds true.

One of the major difficulties in implementing any software is knowing where the data is, how to get hold of it and ensure its robustness, quality and authenticity. Basic principles of data management are available through standard texts. Data must be treated as an independent asset in its own right. It has a cost of creation, a value and a life cycle. Simple data maintenance and management principals are available, and common in most organizations. Data may be efficiently stored in a database. However, the same data could also be inefficiently stored in a different database. The production of a database that will be efficient can be effected using common rules for entity relationship modeling and normalization. However, with complex databases relating many individual entities, the process becomes far more of an art than a science. Development of a data model which would be relevant to both utilities, vendors, suppliers and software contractors would remove the need for redefinition of data bases each time an application is developed. Object-oriented approaches have developed to reflect the close association of data and software – encapsulation, polymorphism and inheritance help to make sure that software performs desired functions only on specific data. As a means of developing software, object orientation provides a firm foundation. They do not, however, help to define a data base structure or format, and do not per se ensure data consistency.

Legacy Systems Where a conversion is not possible, or is not desirable, wrappers (a.k.a. adapters) can be used to communicate to a non-CIM compliant database. This requires conversion expertise and effort, but would allow the legacy system to be maintained efficiently within the data layer.

CIM – the Common Information Model The Common Information Model (CIM) is an abstract model that represents all the major objects in an electric utility enterprise information model [2]. A key purpose of the CIM is to provide a basis for a common language for describing exactly what data is being exchanged among business functions. The CIM is a means of specifying a message in a common format such that it means the same thing to the sender and the receiver. It removes ambiguity regarding nomenclature. The messages are related through a CIM model that provides a hierarchy and a set of terms that describe power system devices. The CIM does not specify how to put a database together, and several CIM compliant databases could be formed for one set of messages. The CIM is still being developed, and is growing to cover elements of the power system in more detail, and to add elements that are of interest to, say, maintenance engineers. Such items would include bushings and oil, say. The CIM has grown out of the need for power system operators to work with each other. Power system models and data have to be transferred between companies. Initially, the EMS vendors seemed very much opposed to this effort – possibly because they could see an end to lucrative lock in contracts for individual companies, requiring mass customization and detailed close collaboration.

Data Models A data model goes beyond the present CIM and allows a database to be built that could be CIM compliant. The data model relates strongly to the CIM, but is needed to allow a common database structure. This approach would allow any organization to build a CIM compliant database, and has been pioneered by some companies who offer CIM Installer software to build such a database.

Integration Bus This provides a means for accessing data in a simple manner. Rather than point to point links between applications and data, as is presently the case, data is ‘published’ via the integration bus. An application that accesses the bus is able to ‘subscribe’ to the data and receive a copy. In this way, integrity of source data is better managed, only one source version is required for all applications, and individual applications no longer need to be configured with the location of source data. When data moves, it is ‘tracked’ by the bus and remains published and subscribable. Clearly, the bus is more than a standard data bus, coming with a lot of intelligence and sophistication to manage data assets. A ‘subscriber’ is a consumer of information. A ‘publisher’ is the source of data. Upon data change, the publisher sends the information to the bus; the bus delivers the new data to consumers. Consumers of information do not need to poll for information that has changed, it is provided when changed. Publishers do not need to know if, in fact, there are consumers actually using the data – the data gets published anyway. There needs to be a monitoring of data published to see what is actually being used – there’s no point in publishing data just for the sake of it. The integration bus data to be published for general use, and then subscribed to by permitted users; one of the back bones for implementing CCAPI/CIM replacing point-to-point communication between applications with a single communication channel. The presence of an integration bus makes implementation of a CIM compliant database much simpler.

UCA – Utility Communication Architecture This has grown out of a need for substation devices to communicate. Similar in purpose to the CIM initiative, the UCA specifies messaging structures based on object models – GOMSFEE bricks. These, as with CIM, give no base for producing a simple and efficient database – in fact, direct storage of data in GOMSFEE format would be highly inefficient, lead to large quantities of duplicate data and would be expensive to maintain and manage. More information on the UCA can be found from EPRI or from the IEEE [3]

IEC Technical Committee 57 This body is responsible for developing standards for power system control and associated telecommunications. TC 57 consists of a number of working groups (WGs), including, but not limited to: •

WG3 – standards for reliable data acquisition and control on narrow band serial data links

•

WG7 – standards for exchange of real-time operational data between control centers over wide area networks

•

WG9 – data communications over distribution line carrier systems

•

WG10-12 – progressing standards for substations, including feed in from UCA2

• •

WG13 – facilitating integration of applications within a control center, including interactions with external operations in distribution etc; CCAPI fed into this WG WG14 – developing interfaces for distributed management systems; also fed by CCAPI

IEC Technical Working Group 13 WG13 was formed to develop Energy Management Systems (EMS) Application Program Interface (API) standards to facilitate the integration: •

EMS systems developed independently by different vendors

•

between entire EMS systems developed independently

•

between an EMS system and other systems concerned with different aspects of power system operation, such as generation or distribution management.

This is accomplished by defining standard APIs to enable these applications or systems to access public data and exchange information independently of how such data is represented internally. These standards are built around a common information model, CIM, which provides an abstract model for a complete power system using Unified Modeling Language (UML). The CIM is part of the overall EMS-API framework: the CIM specifies the semantics for this API. IEC Technical Working Group 14 This WG addresses the need for standards for system interfaces for distribution management systems (SIDMS). WG14 is basing their work on the same CIM as WG13, with extensions specifically needed for distribution systems addressed in WG14.

Drivers With increasing outsourcing of IT from large organizations, there is a benefit in reducing reliance on corporate IT departments for management and maintenance of data and software. Through use of browsable software deployed over the corporate intranet, this reliance may be reduced. Costs for introducing new EPRI software are thus reduced and barriers to entry lowered. Other industries, notably the telecommunication companies, have benefited both from the three-layer approach and from working to line up their ‘data ducks’ [4,5]. Such organizations can gain

competitive advantage by being able to apply software systems to key data in a timely manner. Such business beneficial software is available through the corporate intranet, or through a web portal. The growth in portal technology, both within an organization and over the internet are areas where an organization can further reduce IT overheads, while increasing corporate awareness. Software development practices have grown to include the three-layer architecture [6]. Meanwhile, the development of software itself has grown from extensive specifications and detailed prior analysis into methods where software responds to the turbulent business and technology environment [7]. So called ‘agile software development’ includes such techniques as extreme programming, lean development and adaptive software development. These are not esoteric techniques, but are grounded in a simple premise that satisfying the customer at the time of delivery, rather than at project initiation, takes precedence. Changes in requirements, scope and technology may occur during the project lifespan, and may be outside of the control of the development team. However, rather than planning into minute detail, and analyzing every contingency, build in flexibility to the plan such that any project rework is reduced in impact and reduced in cost. Key to this is close liaison with customers – the first delivery in weeks to achieve an early win and provide rapid feedback. Second is the need for simple solutions which need less subsequent reworking, can be continually improved and continually tested.

Achieving the vision There are several elements that need to be in place to achieve the vision:

Element One - The Three-layer Architecture One way of achieving this vision is to pursue what is known as the three-layer architecture for software. This separates out data from algorithms from presentation. It is a simple and common sense approach. It does require that an organization knows where its data is, and who is responsible for it. The benefit of this approach is that management of data can be separated out from management of applications that act on that data. The value inherent in the data is then quantifiable. Best of breed applications are then able to use that data for business benefit relatively simply, without requiring data duplication or proprietary data formats.

Element Two – Sound Data Management Good data management is a requirement for any organization, not just those in the electricity supply industry. Separating out data and software is a common sense step. It may take some effort, but leads to more flexible software. Separating out the data means that it can be managed more efficiently and effectively, with costs and value ascribed to data. Similarly, software applications can be upgraded without having to re-enter data. A good example is given by NASA space probes to the edges of the solar system. They are limited in the type of sensors they can carry to provide data, but the algorithms that work on the data are updated over the years as improved and more efficient versions are produced. Clearly local intervention is impossible, as there are no local maintenance crews; everything is done remotely.

The IEEE Computer Applications in Power, July 2001 issue [8], notes the need for a globally accepted data model that would benefit individual utilities. This data model would be one that would guide database implementations. This has been investigated by several individual utilities, but it is an area where collaboration would yield benefits. A common data model would help promote common data standards. This is independent of work in extending the Common Information Model (CIM). The CIM specifies a common messaging format for exchange of data, but does not give guidance on efficient and simple database structures to store that data. That is a separate issue. Some point data may be transferred to a warehouse [1,8]. This approach allows data to be mined using tools that provide multi-dimensional analysis. Setting up a data warehouse for this application is an activity that needs time and effort – the most productive approaches are iterative, building on known capabilities and successes [4]. Data warehousing within utilities should be pursued on the basis of common warehouse architectures. Development of such architectures should be within the remit of EPRI.

Element Three – Algorithms Encapsulated By separating out the data, R&D, say, is freed to concentrate on algorithms which provide value. Delivery of those algorithms in software packages is made easier through a standardized approach to data communication and results presentation. Products of software development are algorithms and applications of benefit to electric supply utilities. The vision here is that algorithms are encapsulated in software in such a way that they can be used either as part of a stand alone application, or through software calls from a separate application.

Element Four – Intranet Delivery of Software Making best use of present corporate infrastructures – viz. the intranets which are already in place. Software can be made available through standard browsers, reducing expensive version control and upgrade costs. Key performance indicators, produced through the software, would be published to web pages. The growing field of knowledge management is being both accelerated in its development and underpinned by intranets [5]. It is sensible to make best use of available technology and insist that software be delivered through this medium. As software will also be in a three-layer format, future replacement of the intranet with some other presentation means can be incorporated more rapidly and with less disturbance to available software algorithms. In setting up software to be intranet compatible, it is sensible to make sure that it is also easily transformed into internet deliverable software. This requires a level of expertise, investment of time and planning at an early stage to offset much larger requirements for time and resources ‘down the line’. The possibility here is of buying a license to run software over the internet, rather than having it installed on the intranet at all.

Element Five – Rapid Prototyping

As research develops new algorithms and new applications for delivery through software, rapid prototyping of these is required to ensure benefits reach client utilities as soon as possible. There is a documented ratio of 10:1 in the efficiency of good software developers to that of poor developers [11]. By standardizing in a number of areas – such as data formats, messaging, use of intranet, look and feel of presentation layers – the need to re-invent code is reduced, with a consequent impact on the amount of work developers have to do. This means software is more likely to work, and more likely to be available quickly. As the emphasis is on rapid development of prototypes for subsequent commercialization, team work is needed, bringing together developers and users to work on and try out real working software [7]. This is preferable to development of comprehensive documentation a priori for a product that may have few users in a particular form – it is better to document a successful product rather than try to ‘sell’ a well-documented product. The ‘MMC’ approach is relevant: look at what the software must provide (M), what it must not do (M) and what it could do (C) [7]. This approach concentrates on the development of what is at the heart of the software, leaving the extras to a later date, and avoiding the bane of creeping elegance.

Benefits Through use of the three-layer approach, utilities will have access to the latest applications from EPRI, the best and most cost effective computer solutions, and ability to deploy best of breed applications from a number of sources in a coherent and simple manner. By working to a common standard and using a common approach for software development, individual applications can be built much more rapidly – they don’t have to start from a blank sheet of paper, and will not have to re-invent the wheel. In addition, developed software will be able to move much more quickly into the commercial arena. User groups will be able to focus on the algorithms, rather than the applications that deliver those algorithms and their supposed benefits. User groups will also be able to discuss the value inherent in different data sources, and the best tools for extracting value from that data. Following this strategy will enable the following benefits: •

Software development and rapid prototyping can be achieved; software developers will be able to work to a simple and robust framework where much basic work in presentation and data structure has already been defined; by requiring algorithms work with CIM/XML software interoperability can be ensured

•

Version control is easier; instead of distributing software to users within the organization, entailing detailed records of who has which version, a single version of an application is made available on the intranet, allowing simple control of version, access and use

•

Data management is easier; data as a separate asset can be maintained and managed more easily if it is in a single place; access to data, ability to create, read, update and delete data can be controlled for different users;

•

Data does not have to be duplicated for particular applications; this reduces the possibility of inconsistency between databases; there is an issue here regarding particular organizations attitude towards data and disparate databases being available to particular users – but use of a three-layer

architecture and integration bus helps to identify where data is, and which applications have access to it •

IT infrastructure involvement in new software deployment is reduced; through single point of installation on the corporate intranet and a standard approach to EPRI software so that having installed one application, further applications with similar EPRI look and feel are easier to install

•

New software with new functionality can be deployed rapidly; a standard look and feel, and simpler access to data sources, new software can be deployed relatively quickly as ‘one more button on the favorites list’

•

More users will be able to access the same data, using the same software; costs of deployment per head are thus reduced

•

There will be less one off fixes at individual organizations, less customization on site, fewer consultants playing with infrastructure to make software work rather than concentrating on delivery of software benefit

•

There will be a reduction in data management overheads; an initial cost in providing the right framework will be followed by reduced maintenance and management costs

•

Software can be supplied in a standard format by a number of vendors, increasing competition and reducing the chance of supplier ‘lock in’

•

Overall, with an intranet infrastructure, software maintenance and management costs are reduced

References [1] C. J. Date, “An Introduction to Database Systems”, 7th Edition, Addison Wesley, October 1999 [2] Common Information Model (CIM) : CIM 10 Version1001976 - 11/14/2001 - 01T039.0h [3] CAP Tutorial: “Fundamentals of Utilities Communication Architecture”, IEEE Computer Applications in Power, July 2001 [4] T. Petrik, “Utilities Boost IT Spending as Confidence in Technology Grows”, Electric Light and Power, April 2001 [5] C. Symons, “Are the Latest second generation Intranets the Way Forward in Knowledge Management?”, Managing Information, December 2000 [6] P. Chang, “Knowledge is Power”, Power Engineering International, March 2001 [7] J. Highsmith, A. Cockburn, “Agile Software Development: the Business of Innovation”, IEEE Computer, September 2001

[8] D. Shi, Y. Lee, X. Duan, Q. Wu, “Power System Data Warehouses”, IEEE Computer Applications in Power, July 2001 [9] A. Zeichick, “How Data Mining makes Businesses Smarter”, Red Herring, October 2001 [10] K. Wiseth, “Find Meaning”, Oracle Magazine, September/October 2001 [11] D. Spinellis, “Fear of Coding and How to Reduce It”, IEEE Computer, August 2001

Integrated Monitoring and Diagnostics at Oklahoma Gas and Electric

John E. Maio and Mark J. Danna Data Systems and Solutions

Lonnie Carlon and Jack Hammers Oklahoma Gas and Electric

ABSTRACT Title: Integrated Monitoring & Diagnostics at Oklahoma Gas & Electric Project Start: Fourth Quarter 1998 Technical Expertise: Substation Automation Parties Involved: DS&S, OG&E, Schweitzer and EPRI Authors: John Maio and Mark Danna, Data Systems and Solutions Lonnie Carlon and Jack Hammers, Oklahoma Gas and Electric The Integrated Monitoring & Diagnostics (IMD) system at OG&E is the first EPRI IMD system to be developed, tested, and installed. Integrating substation off-the-shelf IED's with Maintenance Management Workstation (MMW) algorithm capability yielded a cost effective and unique method for Monitoring, Diagnosing, and submitting SAP notifications for suspect failing Substation equipment. When IMD was conceived, the Maintenance Management Workstation (MMW) was in its infancy. MMW was still in Beta testing at all of the Steering Committee sites (i.e. Texas Utilities, Consolidated Edison, Commonwealth Edison, PEPCO and Duke Energy). EPRI along with DS&S realized that the analysis capabilities required in the substation by IMD were similar to the analysis capabilities required by engineers to analyze data across multiple enterprise level database systems. MMW was conceived to provide data analysis at the enterprise level. By adding a database that captures data at the substation level and placing MMW on top of this local substation database repository, all of the requirements of IMD were met. As a result, notifications are issued to Computerized Maintenance Management Systems (CMMS) such as

Maximo, MP2, Engica, and SAP for investigation by substation engineers. For OGE, a special interface was developed to issue notifications into their SAP system. The IMD project goal is simple in some respects. The goal of IMD is to provide “smart” notifications to maintenance engineers on impending problems on substation equipment. The system is a localized system in the sense that it operates in the substation by analyzing data being collected by substation equipment monitoring devices. In addition, the IMD system is distributed because it is installed at each substation that requires monitoring. Hence, the name integrated monitoring & diagnostics is the development of a localized, distributed monitoring system that depends upon 1) a substation computer and database, 2) a tool to analyze incoming data, 3) monitoring devices to collect the data, and 4) diagnostics to provide meaningful output to personnel in order for them to act on the diagnosis. After 2 and ½ years and many project delays and technical challenges, we are finally nearing completion. We plan to enter in Site Acceptance Testing during the fourth quarter of 2001 and closeout the project shortly thereafter. This project was executed without using the Utility Communication Architecture (UCA).

1. INTRODUCTION The goal of IMD is to provide “smart” notifications to maintenance engineers on impending problems on substation equipment. The system was a localized system in the sense that it operated by analyzing data being collected by equipment monitoring devices found in the substation. Hence, the name integrated monitoring & diagnostics was the development of a localized monitoring system that depended upon 1) a substation database, 2) a tool to analyze incoming data, 3) monitoring devices to collect the data, and 4) diagnostics to provide meaningful output to personnel in order for them to act on the diagnosis. When IMD was conceived, MMW was in its infancy. MMW was still in Beta testing at all of the Steering Committee sites. EPRI along with DS&S realized that the analysis capabilities required in the substation by IMD were similar to the analysis capabilities required by engineers to analyze data across multiple enterprise level database systems. MMW was conceived to provide data analysis at the enterprise level. By adding a database which captures data at the substation level and placing MMW on top of this local substation database repository, most of the requirements of IMD were met. The IMD system at OG&E is the first IMD system to be installed. We are finally nearing completion.

Figure 1.0-1 Transformers and Breakers being monitored at the Carnall Substation

2.

IMD COMPONENTS

The three main pieces of the IMD software include: 1) the Maintenance Management Workstation (MMW), 2) a computerized maintenance management system (CMMS), and 3) the Intellution iFix/iCore database. All three software packages are off-the-shelf systems that were modified/configured/integrated for the IMD project.

Figure 2.0-1 IMD System Configuration

The Performance Assessment Module (PAM), within MMW, provides historical data analysis capability through 1) graphing, 2) equations, 3) visual basic scripting, 4) post-processing dynamic link libraries, and 5) many other features necessary to analyze and process operation data collected by a system like IMD. More importantly, PAM uses its analysis capabilities to generate notifications of impending equipment problems. Notifications are sent to humans via email or via application programmer’s interfaces (APIs) to existing data systems. On the OG&E IMD project, PAM provides a notification interface (API) to send problem notifications to the OG&E’s SAP system. The notifications appear in SAP and are processed by humans. For the OG&E project, PAM’s notification interface was modified to support notifications to SAP. In other projects, PAM was modified to support notifications and/or work order generation. CMMS

systems that PAM supports notifications for, includes but is not limited to Maximo, Engica, and MP2. The notification modifications created for this project can be applied to future IMD projects that do not use SAP. The notifications in MMW are generated in a multi-step fashion. The first step is for the Maintenance Ranking Assessor to execute a series of algorithms using data collected in the iFix/iCore database. The algorithms developed in the EPRI Working Groups on Transformers and Circuit Breakers were used to the fullest extent possible. The tool uses equipment algorithms applied through MMW’s equation editor AND data collected by the monitoring devices to calculate equipment condition. MRA normalizes the equipment condition and provides an overall ranking or priority. The data from the monitoring devices is concentrated through the Schweitzer 2030. The 2030 maps data collection streams from the monitoring devices to variables within a database. The iFix database is an off-the-shelf product used to collect data from the monitoring devices via the 2030. OLE for Process Controls (OPC) is a standard interface protocol that eliminates the necessity for vendors working with intelligent electronic devices (i.e. IED) to communicate to these devices via DNP3.0 or through the many other communication protocols currently found in the market. The bottom line is that there are many communication protocols used by IED vendors. OPC provides 1 communication method independent of the protocol. The importance of OPC to the OG&E project became less significant since OPC is only being used as the communication protocol between the iFix database and the Schweitzer 2030. The project originally planned to communicate directly to the individual devices using OPC; however, the IED vendors (e.g. Harley, Hydran, etc.) were not ready to support OPC communication. Since 1998, some of the IED vendors are now committed to supporting OPC, but it is a very slow transition. Schweitzer has commented that they feel the IED vendors are better poised to make this transition to OPC and they would expect the complete transition in the next 2 to 3 years. Because of this change in strategic direction, all of the communication to the IED's occurred through the 2030. Interfacing the 2030 to the IEDs was an arduous process and the elimination of the 2030 in future IMD projects is expected. However, this will only occur when all IED vendors support OPC. There was nothing simple about interfacing the 2030 to the different monitoring devices. The IMD OG&E project has produced 2030 scripting language that will hopefully make this integration easier at other sites. The scripting language is specific to the 2030 and the specific monitoring devices used on this project. Thus, if other monitoring devices exist, 2030 scripting language must be developed specific to those devices.

3. IMD HARDWARE

The IMD hardware for the OG&E project consisted of 1) the Schweitzer 2030, 2) existing substation equipment, and 3) intelligent electronic devices. The existing substation equipment may include previously installed temperature sensors, tap changer indicators, and relays. Note that the Master SEL-2030 is also connected to OG&E’s SCADA system via a Harris D-20 RTU.

Figure 3.0-1 Carnall Substation IMD Diagram

The SEL-2030 is used to concentrate the data. The SEL-2030 is the main message processor that communicates with each of the IED's, retrieves the IED log or current values, and concentrates the data into a single channel. DNP 3.0 is then used to transmit the data via a serial channel to the OPC driver in the controlling PC for historization and processing. In other words, all of the IED's and relays are wired to the SEL-2030. A mapping was created inside the SEL-2030, which maps the incoming data streams to variables within the database. This data is

then formatted using the DNP 3.0 protocol, transferred to the processing PC where it is interfaced with the OPC Driver. The OPC driver is then used to pull values from each point and in turn the tag value is updated inside the iFix software. iFix then takes this data and stores it into a built-in PI database called the Advanced Historian. Once the data is stored, MMW is used to evaluate the data and issue Work Orders or Notifications. The monitoring devices included in the substation consist of the Harley LTC-Map 1525, the Incon OPTImizer+, the Hydran 201Ti and SEL Relays. The monitoring devices are connected to the SEL2030 via RS-232 (serial) and Fiber Transceivers are used as the physical layer. The SEL-2030 is connected to the Processing PC via RS-232 and transmitted via Radio. The Processing PC is located about 2 miles from the substation but could be located at the substation. The Processing PC was tied to OG&E’s Wide Area Network for access to SAP and remote access to the system. Interfacing each vendors IED with the SEL-2030 was perhaps the most challenging aspect of the IMD program. Although each vendor provides a software tool for monitoring the IED, the goal of the IMD project was to pull data directly from the IED into the SEL-2030. Each vendors IED presented a challenge when interfacing the software with the SEL-2030. The documentation for the IED’s omitted key details that were essential for retrieving and interpreting the data. In one case, the IED EEPROM or said differently, the IED's internal program, had to be modified to present the data in such a fashion that it could be translated to the DNP 3.0 Map. In other case, the instruction code for retrieving the data from the IED was so terse, that it hardly made sense. Had each IED been OPC compliant, integration would have been much easier. The common element that allowed us to overcome the interface obstacle was that each device did provide an RS232 interface and instruction sets for retrieving the internal log files and values. The programming flexibility of the SEL-2030 was also key in integrating non-industry standard devices. With the advent of the UCA 2.0 interface, perhaps the IED's could be made to speak a common language. The Harley LTC-MAP 1525 is used for monitoring multiple aspects of the Transformer Tap Changer, the Transformer top oil and main tank temperatures, and provides auxiliary inputs for assimilating data from other 4-20ma output devices. With CT's attached to tap changer motor supply and with inputs for the transformers output voltage and current, the tap changer start/stop cycle is evaluated and data is logged.

Figure 3.0-2 Harley LTC-Map 1525 Installed in Tap Changer Cabinet.

Figure 3.0-3 Harley LTC-Map 1525 and Incon Tap Changer Position Indicator installed in Tap Changer Cabinet.

Figure 3.0-4 Transformer Temperature Sensors on Tap Changer Tank and Transformer Main Tank

The Incon OPTImizer+ is used to monitor breaker operations. For each breaker operation, the contact arc time and breaker operation time are recorded. "I squared T" (I2T) values for monitoring and diagnosing breaker operations are stored in a log file that is retrieved by the SEL-2030. Note that the Incon EEPROM was customized to work with the IMD system. A delay entry was added so that the log could be download about once a minute without loosing any data off the log entry list.

Figure 3.0-5 Close-up view of an installed Incon OPTImizer.

Figure 3.0-6 Incon mounted inside Breaker Cabinet.

Figure 3.0-7 External view of Breaker.

It is important to note that the IED's, including the Harley and the OPTImizer+, have built in algorithms for calculating data in a real-time mode based upon internal algorithms. The resultant data is then logged for determining the breaker or transformer health. The advantage of using an IED in a real-time mode versus just monitoring the output voltages and current is an important key for determining the substation health. Moreover, the IED's present the critical information that has been consolidated and pre-processed for further evaluation as a much smaller data set.

The Hydran 201Ti is used for monitoring the dissolved gases in the transformer oil. The temperature of the transformer oil and the temperature of the Hydrogen sensor in combination of the Hydrogen level are all indicators of the health of the transformer. Note that the Hydran 201Ti can be interfaced either via RS-232 directly into the SEL-2030 or the H2 levels can be returned via in input channel on the Harley LTC-Map.

Figure 3.0-8 Hydran 201I mounted on transformer.

Figure 3.0-9 Hydran 201i mounted on transformer.

The SEL Relays are interfaced with the SEL-2030 and the data is historized with the iFix software. At this time, the relay data is being used for monitoring when the fault current exceeds a preset value, for viewing real and reactive power, and for viewing phase currents and voltages.

Figure 3.0-10 SEL-2030 Master and Slave units in Carnall Substation Building.

Figure 3.0-11 Front view of SEL relays, 2030’s and Fiber Transceivers at Carnall Substation

Figure 3.0-12 Rear view of SEL relays, 2030’s, and Fiber Transceivers at Carnall Substation.

Note that the integration of the IED's was a challenging goal for many reasons including but limited to: 1) The largest number that the DNP 3.0 map can handle is approx. 32K. Therefore, scaling of values in the SEL-2030 before formatting the DNP 3.0 map was required. On the iFIX or receiving end, the values had to be scaled back to their original value. 2) Calibration of each of the IED's inputs is required for accurate data. 3) Poor documentation greatly extended the implementation time. 4) Limitations of the IED's, the SEL-2030, and the DNP 3.0 map were overcome by working with the IED vendors leading to IED EEPROM modifications in one case and using alternate preprocessed IED values for determining the health of the substation equipment.

4. EPRI ALGORITHMS The algorithms are the heart of the IMD system. They are the equations that are used to monitor and diagnose for hardware failures. The breaker algorithms monitor the substation breakers for excessive operating speed, arc duration, fault limits, and restrikes. The transformer algorithms monitor the substation transformers and associated tap changer. They monitor for excessive transformer oil temperatures, comparative temperature evaluations between the tap changer oil temperature and the main tank temperature, tap changer malfunctions, dissolved gas in the transformer oil, and for monitoring when phase currents are exceeded.

The relay algorithm is currently only monitoring the fault current, but note that all relay data is returned to iFix for monitoring and for future algorithms. To ensure for greater flexibility in the future, all data generated by each of the IED's is returned to iFix and historized. This ensures that future algorithms or monitoring of conditions is easily obtainable. EPRI formed working groups to create the algorithms used for monitoring and diagnosing the substation equipment. Although the algorithms are an extensive compilation there was some disconnect between working groups, equations developed in the working group documents, and “real-life” equations. DS&S, upon implementing the algorithms, found that the algorithms had many problems and did not necessarily conform to any particular IED vendor. In other words, a one sided view of the equation was created. The working groups did not analyze the possible variables generated by the IED vendors in order to determine whether or not the data actually existed. A table listing the outputs available from each vendors IED would have been helpful when incorporating the algorithms. The algorithms presuppose that that data would be available but gave no clarification about how to obtain the data. Having a tractability matrix between equation variables and existing IED variables will assist future IMD projects.

5. CONCLUSION The project encountered many hurdles and walls that caused delays and changes to the overall strategic direction, respectively. It was surprising to have some very fundamental problems in communicating with the IED’s. It almost seems as if nobody has traversed the path of communicating and retrieving data from the aforementioned IEDs. Many of the problems with the communication were due to poor or no documentation while other problems resulted in modifications to fix software/hardware interface problems within the IEDs. Because of these fundamental problems, it leads one to believe that a project of this type using the aforementioned IEDs has never really been performed. The original vision of the project was for the IED vendors to create OPC interfaces. Because the vendors did not participate with the implementation of OPC, the arduous task of communication with the IEDs was implemented in the Schwietzer 2030. It is strongly believed that OPC would have saved a substantive amount of time due to the alleviation of programming IED specific commands (proprietary language) within the 2030 to communicate with each IED vendor using each vendor’s proprietary language. However, given the problems encountered with the IED vendors, we can not be confident that talking to the devices using OPC and/or UCA would have taken any less amount of time.

5 CLOSING SESSION Panel Members Ray Lings, EPRI, Chair Donny Helm Rose, TXU Terry Jackson, Allegheny Power System Don Rose, TXU Bjorn Holm, International Switchgear Consulting Ltd Michael Ingram, Tennessee Valley Authority The Closing Session panel provided an opportunity for the session chairs to present their comments about the papers previously presented and to provide additional comments. The following summarizes their discussions: Approximately 70% of the papers that were presented this conference are based on EPRI research. Many of the projects that have been described are the products of years of research. An integral component of EPRI research is the input gained at the various working group or task force meetings. These are meetings where EPRI meets with the utility engineers, not the management of the companies, and the contractors or researchers to shape out the projects. A minimum of two meetings per year are held for each project. True value is added to EPRI research at these meetings and EPRI members are urged to actively participate in them. EPRI Solutions can get the results of the research out into the field to get applied and used. Once the results of a research project have been demonstrated, it is turned over to EPRI Solutions to take it further and to apply it. Life Extension Guidelines is an EPRI project that presents the best practices for substation equipment based on completed research. It is now an 800 page document and is something like a bible for substation management. It is completely revised every two to three years. Also, EPRI Solutions has a course that covers the Life Extension Guidelines. The fundamental research on UCA has been completed. AEP has set up a test bed for use by manufacturers to confirm UCA compliance for their devices. The whole issue of sharing information, from the substation to the control building to the head office of the utility, needs an organization like EPRI to take a leadership role and help accelerate its development.

MMW continues to be a “flagship” project for EPRI. There are now over 15 installations in various stages. Research is continuing on the algorithms that describe the performance of transformers and circuit breakers. The integration of CIM, UCA and MMW and the movement of asset management data around the utility is of core importance. In addition, the integration of various software packages, that are now stand-alone, can be brought about through CIM. Sharing of equipment failure information will allow better decision making when it comes to maintenance. Also, such information can become the basis for industry benchmarks. Economic analysis must be brought into some of the technical programs. How do you make the decision whether to run a transformer “hot”? Doing so may take some life out of the transformer, but you may be able to make some money in the market. The papers presented at this conference have shown that the utilities are using the programs and methodologies developed by EPRI. Utilities are also getting a better handle on cost/benefit analysis for using these programs. It isn’t easy to do, but there have been some examples presented at this conference on how these programs can help. The programs can also be used as a teaching tool for new engineers. The manufacturers need to be challenged to adapt technologies used by other industries to our industry. For example, technologies used in the medical or mechanical fields could have application to us. More can be gained through gas in oil analysis for circuit breakers. Moisture is also an important factor. A surprising amount of interest was noted about the paper on cell management for station batteries. The message to the attendees was quite successful. The need for power quality keeps going up, O&M budgets are going down, money for infrastructure improvements are disappearing and maintenance outages are harder to get. Monitoring of substation equipment can assist us in determining if equipment needs to be maintained. We need to educate others (neighbors, bosses, and regulators) about the capabilities of the systems that we are designing and building and that there is a cost and a benefit associated with it. They need to understand how the system fits together. The use of infrared thermography is very important in being able to identify equipment that is operating improperly. Field information must be accurate and sensing equipment must be properly calibrated.

PT Load was used by Allegheny Power to operate through a crisis last summer. Good use was made by them with this tool that was developed by EPRI. Acoustics and DGA are coming together. The paper by Waters presented a correlation between gas generation and acoustical emissions. The Aubin paper presented a valuable analytical model for evaluating the cost effectiveness of monitoring equipment. The Schellhase paper demonstrated what happens inside an LTC as it naturally operates. Different products are produced, none of which are helpful to the operation of the LTC. The paper demonstrated the effectiveness of on-line oil filtration for an LTC. Data is being gathered and we are all trying to get to the “information age”. The problem is that we haven’t reached where we want to be, and that is to the wisdom part of it. How do you integrate all this information together so it tells us something? One of the hurdles is the lack of standards. We need to push the development of standards between organizations. A common model is needed to work with. You can’t jump from data to wisdom without these standards in place. Part of the problem is internal to ourselves— the data islands that exist that need to be brought together. Difficulties remain getting the information we need from different devices because of the lack of standards. The need for standards cannot be over-emphasized. Also, not all technologies are being used to their full capability. IED information needs to be brought together so that it can be converted into information to make business decisions and to optimize the performance of our equipment. We have come a long way, but we still have a ways to go. The technical side of out industry needs to have a bigger voice in the financial side in order to continue to move the infrastructure forward and to implement the type of systems that are key to the future of the industry. At this time, productivity improvements are not emphasized as much as cost reduction. BC Hydro now does DGA testing on LTCs as a result of research sponsored by EPRI. The benefits of doing this can be stated in dollars. We need to re-create the link with the educators. There is a void in our industry. There are opportunities and enjoyment in this work that we have to communicate to them on why they should want to come into our industry.

6 APPENDIX: LIST OF ATTENDEES

Report ID:

A75C2254

Page No. Run Date Run Time

LIST OF REGISTRANTS BY NAME

2/17/02

Name/Title/Company

Substation Equipment Diagnostics Conference through 2/20/02 Catalog Number:

Address

Abramov,Gary Manager Graviton, Inc.

858-909-2172 (phone) 9820 Town Centre Dr San Diego, CA 92121

Acevedo,Javier Enrique Electrical Engineer Interconexion Electrica S.A.

Calle 12 Sur No. 18-168 Medellin 33405 , Colombia

175 E Old Country Rd Hicksville, NY 11801-4280

1610 Matthews Ave Bronx, NY 10462-3999

5411 NE Highway 99 Vancouver, WA 98663-1302

4651 S Westmoreland Rd Dallas, TX 75237-1017

214-330-7379 (fax) [email protected] 514-369-5744 (phone)

179 Brunswick Blvd Point Claire, QC H9R 5N2 , Canada

Avery,Larry D. Vice President, Engrg. & Operations Alabama Electric Cooperative, Inc.

360-418-2958 (fax) [email protected] 214-330-3245 (phone)

Aubin,Jacques Transformer Specialist GE Energy Services

718-828-8823 (fax) [email protected] 360-418-2652 (phone)

Anand,Ramesh

AVO International, Inc.

516-545-3662 (fax) [email protected] 718-315-5484 (phone)

An,Sokom

Bonneville Power Administration (BPA)

574-309-3414 (fax) [email protected] 516-545-4359 (phone)

Alfieri,Michael Manager, Substation Operations Consolidated Edison/O&R

[email protected] 574-309-4455 (phone)

Adelson,Polina Lead Engineer, System Planning Long Island Power Authority

Phone/Fax/Email

514-694-9245 (fax) [email protected] 334-427-3284 (phone)

Highway 29 N Andalusia, AL 36420

334-222-7785 (fax) [email protected]

1 06.Mar.2002 12:30:46PM

E00000000000209138

Banach,Zbigniew

M/S 8th Fl.

630-576-6989 (phone)

Sr. Engineer Exelon Corporation

2 Lincoln Ctr Oakbrook Terrace, IL 60181-4295

630-576-6355 (fax) [email protected]

Baranczyk,Randy

608-787-1253 (phone)

Manager, Transmission Maintenance Dairyland Power Cooperative

3200 East Ave S La Crosse, WI 54601-7291

608-787-1475 (fax) [email protected]

Bell,Sandra C.

M/S MR 2T

423-751-4953 (phone)

Project Engineer, RSO&E Tennessee Valley Authority (TVA)

1101 Market St Chattanooga, TN 37402-2801

423-751-2463 (fax) [email protected]

Benach,Jeff Project Engineer AVO International, Inc.

610-676-8513 (phone) PO BOX 9007 Valley Forge, PA 19485-1007

Benavidez,Martin J.

City Public Service, San Antonio

210-353-4624 (phone) 145 Navarro St San Antonio, TX 78205-2986

Berube,Richard

Morgan Schaffer Corp.

5110 Courtral Avenue Montreal, QC H3W 1A7 , Canada

145 Navarro St San Antonio, TX 78205-2986

8205 Kensington Blvd Ste 640 Davison, MI 48423

Brown,Richard

810-654-0415 (fax) [email protected] 416-207-6874 (phone)

800 Kipling Ave Toronto, ON M8Z 6C4 , Canada

Brown,Christopher Business Development Manager Foster-Miller, Inc.

210-353-3340 (fax) [email protected] 810-654-0415 (phone)

Braun,Jean-Marie Nuclear Consultant Kinectrics, Inc.

514-739-0434 (fax)

210-353-3157 (phone)

Braschayko,Anne Exhibit Manager Carpentier Events

[email protected] 514-739-1967 (phone)

Boston III,James A. T&D Engineer I/Operations City Public Service, San Antonio

610-676-8584 (fax) [email protected]

416-207-5773 (fax) [email protected] 781-684-4042 (phone)

350 2nd Ave Waltham, MA 02451-1196

781-890-8515 (fax) [email protected] 610-676-8611 (phone)

V.P., Strategic Mgmnt. AVO International, Inc.

PO BOX 9007 Valley Forge, PA 19485-1007

Brucetti,Robert C. Product Manager Doble Engineering Co.

617-393-2929 (phone) 85 Walnut St Watertown, MA 02472-4037

Buckley,Scott J. Operations Research Analyst Potomac Electric Power Co.

3300 Benning Rd NE Washington, DC 20019-1501

1100 W 116th Ave Denver, CO 80234-2814

303-254-6035 (fax) [email protected] 913-967-1757 (phone)

9800 Metcalf Ave Overland Park, KS 66212-2216

Campbell,Todd Substation Supervisor Austin Energy

[email protected] 303-452-6111 (phone) Ext: 6441

Bush,Rick Editor-in-Chief Transmission & Distribution World

617-926-4900 (fax) [email protected] 202-388-2386 (phone)

Buckridge,Dave Transmission I&M Superintendent Tri-State G&T Association, Inc.

610-676-8510 (fax) [email protected]

913-967-1905 (fax) [email protected] 512-505-7058 (phone)

2526 Kramer Ln Austin, TX 78758-4416

Cannon,Mike

512-505-7077 (fax) [email protected] 712-279-8750 (phone)

Cannon Technologies, Inc.

311 Terra Centre Sioux City, IA 51101

712-279-8751 (fax) [email protected]

Cantrelle,Donald

M/S B20033

402-506-3581 (phone)

Senior Engineer Georgia Power Co.

241 Ralph McGill Blvd NE Atlanta, GA 30308-3374

402-506-4079 (fax) [email protected]

Carden,John Camron

M/S SB

770-270-7724 (phone)

Substation Maint. Eng. Georgia Transmission Corp.

2100 E Exchange Pl Tucker, GA 30084-5336

770-270-7575 (fax) [email protected]

Cargol,Tim

Massachusetts Institute of Technology Carlon,Lonnie

617-253-5019 (phone) 155 Massachusetts Ave Cambridge, MA 02139-4205

617-252-1654 (fax) [email protected] 405-553-8132 (phone)

Specialist, Prot. & Contol Tech. OG&E Electric Services

3220 S High Ave Oklahoma City, OK 73129-5030

Chantigny,Denis Regional Sales Manager, Canada GE Energy Services

514-693-1400 (phone) 179 Brunswick Blvd Point Claire, QC H9R 5N2 , Canada

Clare,Ron Manager, Substation Operations Long Island Power Authority

175 E Old Country Rd Hicksville, NY 11801-4280

145 Navarro St San Antonio, TX 78205-2986

23512 NE 19th Dr Sammamish, WA 98074

508-632-0262 (fax) [email protected] 404-675-1883 (phone)

62 Lake Mirror Rd Bldg 3 Forest Park, GA 30297-1613

Cox,Randy Diagnostic Supervisor TXU Electric

210-353-4449 (fax) [email protected] 425-868-2773 (phone)

Coffeen,Larry Research Engineer Georgia Institute of Technology

516-545-5112 (fax) [email protected] 210-353-4386 (phone)

Clinard,Kay President KC Associates

514-694-9245 (fax) [email protected] 516-545-4346 (phone)

Clifton,Darrell A. Project Engineer City Public Service, San Antonio

405-553-8430 (fax) [email protected]

404-675-1885 (fax) [email protected] 972-721-6229 (phone)

115 N Loop 12 Irving, TX 75061-8701

Crutcher,Ernest R.

972-721-6305 (fax) [email protected] 425-885-9419 (phone)

7609 140th Pine Redmond, WA 98052

425-885-9419 (fax)

Microlab Northwest Dahl,William E.

M/S 6W/ED-1

402-636-3031 (phone)

Technical Services Eng. Omaha Public Power District

444 S 16th St Omaha, NE 68102-2247

402-636-3947 (fax) [email protected]

Damsky,Ben

650-855-2385 (phone)

Mgr., Power Elec. Sys. EPRI, Palo Alto

3412 Hillview Ave Palo Alto, CA 94304-1395

650-855-8997 (fax) [email protected]

Danna,Mark

M/S 300

713-346-4239 (phone)

Implementation Specialist

1900 West Loop S

713-346-4201 (fax)

Data Systems & Solutions

Houston, TX 77027-3214

Davydov,Valery

Monash University

[email protected] 39-905-3486 (phone)

Dept. of Elect. & Computer Systems , Eng. Clayton, VIC 3168 , Australia

DeRivi PE,Richard F.

39-905-3454 (fax) [email protected] 925-750-6113 (phone)

Regional Sales Mgr., Western Region GE Energy Services

6130 Stoneridge Mall Rd Ste 275 Pleasanton, CA 94588

925-924-1338 (fax) [email protected]

Degeneff,Robert C.

M/S JEC-7006

518-276-6367 (phone)

Professor, Elec. Power Engineeri Rensselaer Polytechnic Institute

110 8th St Troy, NY 12180-3590

518-276-6226 (fax) [email protected]

Dessureau,Paul

650-855-2185 (phone)

EPRIsolutions, Inc.

3412 Hillview Ave Palo Alto, CA 94304-1395

650-855-8588 (fax) [email protected]

Dighe,Rameshchandra G.

M/S SC-600

512-482-6342 (phone)

Engineer Associate III Lower Colorado River Authority

3505 Montopolis Dr Austin, TX 78744

512-482-6382 (fax) [email protected]

Dominelli,Nick

604-590-7443 (phone)

Dir., Applied Chemistry Powertech Labs., Inc.

12388 88th Ave Surrey, BC V3W 7R7 , Canada

604-590-7489 (fax) [email protected]

Duggan,Patrick M.

M/S 2615-S

212-460-4020 (phone)

Project Manager Consolidated Edison/O&R

4 Irving Pl New York, NY 10003-3598

212-529-8511 (fax) [email protected]

Eckroad,Steven

7-206

650-855-1066 (phone)

Project Manager EPRI, Palo Alto

3412 Hillview Ave Palo Alto, CA 94304-1395

650-855-8997 (fax) [email protected]

Elsey,James L.

440-243-6666 (phone)

Director, New Product Development VRC, Inc.

696 West Bagley Rd Berea, OH 44017

440-826-4646 (fax) [email protected]

Escamilla,Jose H.

M/S 1771

210-353-3317 (phone)

Project Engineer City Public Service, San Antonio

326 W Jones Ave San Antonio, TX 78215-1404

210-353-3340 (fax)

Farias,Richard David

210-353-2914 (phone)

T&D Engineer III City Public Service, San Antonio

145 Navarro St San Antonio, TX 78205-2986

210-353-4449 (fax) [email protected]

Ferreira,Marcos D.

M/S TNSD/AMPN-2

360-418-8631 (phone)

Sr. Electrical Engineer Bonneville Power Administration (BPA)

5411 NE Highway 99 Vancouver, WA 98663-1302

360-418-2958 (fax) [email protected]

Figueroa,Elisa Jarumi

M/S TCT-15

416-345-5167 (phone)

Sr. Network Management Eng. Hydro One Networks, Inc.

483 Bay St Toronto, ON M5G 2P5 , Canada

416-345-5424 (fax) [email protected]

Foata,Marc Research Engineer Institut de Recherche D'Hydro-Quebec

450-652-8306 (phone) 1800 boulevard Lionel-Boulet Varennes, QC J3X 1S1 , Canada

Forrest,George

Uptime Engineered Solutions Co.

450-652-8309 (fax) [email protected] 919-544-8191 (phone)

5317 Highgate Dr Ste 212 Durham, NC 27713-6622

Fox,Will

919-544-2257 (fax)

503-924-3244 (phone)

Director of Marketing SERVERON

3305 NW Aloclek Dr Hillsboro, OR 97127-7101

503-924-3290 (fax) [email protected]

Garrett,Jerald A.

M/S EC13R

704-382-0439 (phone)

Substation Process Specialist Duke Energy Corp.

526 S Church St Charlotte, NC 28202-1802

704-382-6907 (fax) [email protected]

Garza,Jaime

City Public Service, San Antonio

210-353-3506 (phone) 145 Navarro St San Antonio, TX 78205-2986

Gauton,Sean Electrical Power Engineer PowerGen PLC

115-936-2193 (phone) Power Technology Centre , Ratcliffe-on-Soar Nottingham NG11 0EE , United Kingdom

Giesecke,Jon Louis Project Manager EPRIsolutions Lenox Center

210-353-4449 (fax) [email protected]

115-936-2711 (fax) [email protected] 610-518-1615 (phone)

112 Wildbriar Rd Downingtown, PA 19335-1042

[email protected]

Gonzales,Samuel Substation Elec. Supervisor Austin Energy

512-505-7280 (phone) 2526 Kramer Ln Austin, TX 78758-4416

Goodman,Mark Vice President U.E. Systems, Inc.

512-505-7077 (fax) [email protected] 914-592-1220 (phone) Ext: 629

14 Hayes St Elmsford, NY 10523-2536

Gordon,Mary

914-347-2181 (fax) [email protected] 214-330-3536 (phone)

AVO International, Inc.

4271 Bronze Way Dallas, TX 75237-1019

214-331-3937 (fax) [email protected]

Graziano,Joseph A.

M/S MR 2T

423-751-7500 (phone)

Project Manager Tennessee Valley Authority (TVA)

1101 Market St Chattanooga, TN 37402-2801

423-751-2463 (fax) [email protected]

Hagman,Wayne

Room 10-026

617-253-2087 (phone)

Research Engineer Massachusetts Institute of Technology

77 Massachusetts Ave Cambridge, MA 02139-4307

617-258-6774 (fax)

Hanson,Dave General Manager TJ/H2b Analytical Services, Inc.

916-361-7177 (phone) 3123 Fite Cir Sacramento, CA 95827-1816

Harlow,James H.

916-361-7178 (fax) [email protected] 727-593-5185 (phone)

Prinicipal Harlow Engineering Associates

11100 - 141 Way Largo, FL 33774-4434

727-596-2907 (fax) [email protected]

Helm,Donny

M/S ESB-1002

817-215-6170 (phone)

Performance Analysis TXU Electric

PO BOX 970 Fort Worth, TX 76101-0970

[email protected]

Heywood,Richard

National Grid Company plc

13-7238350 (phone) John Forrest House , Kelvin Avenue Leatherhead, SURREY KT22 7ST , United Kingdom

Hinshaw,John Senior Engineer SERVERON Holm,Bjorn

[email protected] 503-924-3244 (phone)

3305 NW Aloclek Dr Hillsboro, OR 97127-7101

503-924-3290 (fax) [email protected] 604-594-4176 (phone)

Consultant Int'l Switchgear Consulting, Ltd.

11078 Upper Canyon Rd Delta, BC V4E 3L9 , Canada

Hurt,Dennis

[email protected] 503-924-3244 (phone)

Vice President, Sales SERVERON

3305 NW Aloclek Dr Hillsboro, OR 97127-7101

503-924-3290 (fax) [email protected]

Ingram,Michael R.

M/S 2T-C

423-751-7799 (phone)

Program Mgr., Transmission Perf. Tennessee Valley Authority (TVA)

1101 Market St Chattanooga, TN 37402-2801

423-751-6087 (fax) [email protected]

Jackson,Terry J. Technical Advisor Allegheny Energy, Inc.

724-838-6486 (phone) 800 Cabin Hill Dr Greensburg, PA 15601-1689

Jakob,Fredi Director of Laboratory Services Weidmann ACTI, Inc.

916-455-2284 (phone) 4011 Power Inn Rd Sacramento, CA 95826-4342

Johnson,Larry Distribution Technology Manager PSEG Energy Technologies

80 Park Plz Newark, NJ 07102-4194

1225 Carnegie St Ste 109 Rolling Meadows, IL 60008

12-046 Waverly Winnipeg, MB R3T 1X6 , Canada

Lane,John

204-475-0273 (fax) [email protected] 907-762-4694 (phone)

5601 Minnesota Dr Anchorage, AK 99518-1081

Knoll,Ernst Research & Development Elin Transformatoren Ges M.B.H.

847-506-1760 (fax) [email protected] 204-474-4962 (phone)

Kimmel,Charlie Manager, Substation Operations Chugach Electric Association, Inc.

973-643-0978 (fax) [email protected] 847-506-1126 (phone)

Kennedy,John

Manitoba Hydro-Electric Board

916-455-0191 (fax) fjakob@weidmann_actI.com 973-430-7919 (phone)

Karlstrom,Mats Regional Sales Manager GE Energy Services

724-830-5020 (fax) [email protected]

907-762-7661 (fax) [email protected] 317-6062784 (phone)

Elingasse 3 Weiz 8160 , Austria

317-6414 (fax)

541-752-7233 (phone)

Sales Manager XENERGY, Digital Inspections

804a NW Buchanan Ave Corvallis, OR 97330-6218

Leinhauser,George T. Engineering Analyst Exelon Corporation

610-648-7924 (phone) 1040 Swedesford Rd Berwyn, PA 19312-1050

Lettmoden,William Senior System Specialist Entergy Nuclear Northeast

541-758-3666 (fax) [email protected]

610-648-7794 (fax) [email protected] 914-736-8345 (phone)

Indian Point Energy Center - IP3 , 295 Broadway Ste 3 Buchanan, NY 10511

Lindgren,Stanley

914-736-5402 (fax) [email protected] 503-924-3219 (phone)

Director, Product Strategy SERVERON

3305 NW Aloclek Dr Hillsboro, OR 97127-7101

503-924-3290 (fax) [email protected]

Lings,Raymond

6-219

650-855-2177 (phone)

Area Mgr., Transmission & Subs. EPRI, Palo Alto

3412 Hillview Ave Palo Alto, CA 94304-1395

650-855-8997 (fax) [email protected]

Loynes,Kenneth Project Engineer EPRIsolutions Lenox Center

413-499-5712 (phone) 115 E New Lenox Rd Lenox, MA 01240-2245

Luby,Tom

[email protected] 405-239-8600 (phone)

Vigilink Corp.

840 Research Pkwy Ste 225 Oklahoma City, OK 74114

[email protected]

Lundberg,Alan D.

7th Fl.

612-230-2889 (phone)

Manager, Substation Maint. Eng. Xcel Energy Services, Inc.

414 Nicollet Mall Minneapolis, MN 55401-4993

612-573-9362 (fax) [email protected]

Mahurin,Michael E. Superintendent City Public Service, San Antonio

210-353-3343 (phone) 145 Navarro St San Antonio, TX 78205-2986

Marquez,Fidel Vice President City Public Service, San Antonio

210-353-3777 (phone) 145 Navarro St San Antonio, TX 78205-2986

Martin,Steve Transmission Manager

210-353-4618 (fax) [email protected]

210-353-4339 (fax) [email protected] 903-872-4647 (phone)

1601 Bryan St

903-872-8251 (fax)

TXU Electric

Dallas, TX 75201-3401

McDermott,Gary Plant Supervisor II U.S. Bureau of Reclamation

970-240-6306 (phone) 1820 S Rio Grande Ave Montrose, CO 81401-4859

McGrail,Anthony J.

National Grid Company plc

John Forrest House , Kelvin Avenue Leatherhead, SURREY KT22 7ST , United Kingdom

3305 NW Aloclek Dr Hillsboro, OR 97127-7101

195 Clarksville Rd Princeton Junction, NJ 08550-5303

3305 NW Aloclek Dr Hillsboro, OR 97127-7101

3305 NW Aloclek Dr Hillsboro, OR 97127-7101

145 Navarro St San Antonio, TX 78205-2986

3412 Hillview Ave Palo Alto, CA 94304-1395

650-855-8588 (fax) [email protected] 215-943-1457 (phone)

9071 Hill Creek Rd Apt 313 Levittown, PA 19054

Parmalee,Jeff Product Manager GE Energy Services

210-353-4449 (fax) [email protected] 650-855-2818 (phone)

Nunez,Arturo Electrical Engineer Quality Services Laboratories

503-924-3290 (fax) [email protected] 210-353-4589 (phone)

Nikolic,Darije Substation Business Dev. Mgr. EPRIsolutions, Inc.

503-924-3290 (fax) [email protected] 503-924-3244 (phone)

Mulhearn,Peter

City Public Service, San Antonio

609-716-0706 (fax) [email protected] 503-924-3200 (phone)

Morgan,Dan Director of Engineering SERVERON

503-924-3290 (fax) [email protected] 609-716-4042 (phone)

Moon,Jim President, CEO SERVERON

137-238-3958 (fax) [email protected] 503-924-3244 (phone)

Miller,Ronnie K. Executive Director, ES&I Physical Acoustics Corporation

970-240-6304 (fax)

137-238-3838 (phone)

Miller,Roger Account Manager SERVERON

[email protected]

609-716-0706 (fax) [email protected] 330-204-0050 (phone)

75 Washington Ave Cuyahaga Falls, OH 44221

419-710-0768 (fax) [email protected]

Payette,Wally

210-260-2857 (phone)

Doble Engineering Co.

2109 Lockhill Selma San Antonio, TX 75213

Perlichek,Robert Lee

M/S 0600

505-241-4807 (phone)

Substation Engineer Public Service Co. of New Mexico

Alvarado Sq Albuquerque, NM 87158-0001

505-241-2363 (fax) [email protected]

Picher,Patrick Researcher Hydro-Quebec

450-652-8256 (phone) 1800 Blvd Lionel Boulet Varennes, QC J3X 1S1 , Canada

Portillo,Homer Engineer Associate C Austin Energy

512-505-7133 (phone) 2526 Kramer Ln Austin, TX 78758-4416

Pressler,Glenn Project Engineer City Public Service, San Antonio

450-652-8051 (fax) [email protected]

512-505-7004 (fax) [email protected] 210-353-2212 (phone)

326 W Jones Ave San Antonio, TX 78215-1404

210-353-3340 (fax) [email protected]

Pullins,Steve

Data Systems & Solutions

5321 Westover Lake Way Hilliard, OH 43026-7217

Raymond,Timothy C. Engineer Power Delivery Consultants, Inc.

518-384-0297 (phone) 28 Lundy Ln Ste 102 Ballston Lake, NY 12019-2107

Rogers,Dan Manager, Suffolk Protection KeySpan Energy

518-399-4051 (fax) [email protected] 516-545-3177 (phone)

175 E Old Country Rd Hicksville, NY 11801-4257

Roizman,Oleg

516-545-5629 (fax) [email protected] 39-905-9612 (phone)

Intell Power

2110 Warinard Carnegie, VIC 3163 , Australia

[email protected]

Rose,Don W.

Room 1010

817-215-6171 (phone)

Mgr., Transmission & Substation Ops TXU Electric

115 W 7th St Fort Worth, TX 76102-7033

817-215-6059 (fax)

Savio,Leo J.

610-662-6103 (phone)

Adapt Corp.

PO BOX 853 Blue Bell, PA 19422-0853

610-279-0797 (fax) [email protected]

Schainker,Robert B.

6-232

650-855-2549 (phone)

Product Line Leader EPRI, Palo Alto

3412 Hillview Ave Palo Alto, CA 94304-1395

650-855-8997 (fax) [email protected]

Schellhase,Hans Research Chemist Powertech Labs., Inc.

604-590-7420 (phone) 12388 88th Ave Surrey, BC V3W 7R7 , Canada

Schikarski,Peter

Hipotronics

845-279-8091 (phone) 1650 Rt 22 Brewster, NY 10509-4013

Schwabe,Robert J. Sr. R&D Engineer New York Power Authority

123 Main St White Plains, NY 10601-3104

1650 Rt 22 Brewster, NY 10509-4013

85 Walnut St Watertown, MA 02472-4037

53 Cedar Creek Rd Placitas, NM 87043-9008

Spare,John H.

505-867-1096 (fax) [email protected] 210-353-2510 (phone)

145 Navarro St San Antonio, TX 78205-2986

Sobocki,Ryszard

Polish Power Grid Co.

617-926-0528 (fax) [email protected] 505-867-1091 (phone) Ext: 5290

Snell,Charles

City Public Service, San Antonio

845-279-8034 (fax) [email protected] 617-926-4900 (phone)

Shumard,Rick Regionl Mgr., South Central EPRI, Palo Alto

914-287-6860 (fax) [email protected] 845-279-8091 (phone)

Short,Scott Regional Sales Manager Doble Engineering Co.

845-279-8034 (fax)

914-287-3794 (phone)

Searls,Brian Director Sales & Marketing Hipotronics

604-590-5347 (fax) [email protected]

210-353-4449 (fax) [email protected] 22-693-2263 (phone)

ul. Mysia 2 Warsaw 00-496 , Poland

22-693-2167 (fax) [email protected] 215-997-4500 (phone) Ext: 287

Senior Principal Consultant KEMA Consulting

4377 County Line Rd Chalfont, PA 18914-1825

Sparling,Brian

GE Energy Services

215-997-3818 (fax) [email protected] 403-214-4448 (phone)

2728 Hopevoell Pl NE Calgary, AB T1Y 9S7 , Canada

Spieth,Don

403-287-9237 (fax) [email protected] 847-506-1126 (phone)

Sales Support Manager GE Energy Services

1225 Carnegie St Ste 109 Rolling Meadows, IL 60008

847-506-1760 (fax) [email protected]

Sublett,Fulton

M/S CP M35

254-897-5661 (phone)

System Engineer TXU Electric

Comanche Peak Power Station , 8 Miles W Of Glen Rose Texas Glen Rose, TX 76043

254-897-0972 (fax) [email protected]

Thaden Jr.,Malcolm V.

Rm. 8064

202-872-2713 (phone)

Principal Engineer, Asset Mgmt. Potomac Electric Power Co.

Edison Place , 701 9th St NW Washington, DC 20058-0001

202-331-6692 (fax) [email protected]

Torres,Hector V. Manager City Public Service, San Antonio

210-353-6454 (phone) 145 Navarro St San Antonio, TX 78205-2986

Traub,Thomas Substation Equipment Consultant Traub Consulting

210-353-4618 (fax) [email protected] 312-335-8842 (phone)

200 E Delaware Pl Apt 30f Chicago, IL 60611-7706

Udstuen,James E.

312-335-8842 (fax) [email protected] 610-676-8568 (phone)

V.P., General Manager AVO International, Inc.

PO BOX 9007 Valley Forge, PA 19485-1007

610-676-8602 (fax) [email protected]

Vagle,Pete M.

Bin 20033

404-506-3915 (phone)

Senior Engineer Georgia Power Co.

241 Ralph McGill Blvd NE Atlanta, GA 30308-3374

404-506-4079 (fax) [email protected]

Van Der Zel,Luke Project Manager EPRI, Palo Alto Vo,Trieu

413-499-5012 (phone) 115A E New Lenox Rd Lenox, MA 01240

[email protected] 210-353-2127 (phone)

T&D Engineer II/Operations City Public Service, San Antonio

145 Navarro St San Antonio, TX 78205-2986

Vujovic,Predrag Director, Engineered Strategies Doble Engineering Co.

617-393-3127 (phone) Marlborough Executive Park , 65 Boston Post Rd W Ste 390 Marlborough, MA 01752

Ward,Barry Manager, Power Transformer EPRI, Palo Alto

210-353-3340 (fax) [email protected]

617-926-0528 (fax) [email protected] 650-855-2717 (phone)

3412 Hillview Ave Palo Alto, CA 94304-1395

Wasylenko,Stephen

650-855-8997 (fax) [email protected] 516-545-3351 (phone)

Associate Engineer KeySpan Energy

175 E Old Country Rd Hicksville, NY 11801-4257

[email protected]

Waters,Thomas

M/S C-100

503-924-3200 (phone)

Senior Chemist SERVERON

3305 NW Aloclek Dr Hillsboro, OR 97127-7101

503-924-3290 (fax) [email protected]

Weiblen,Mike Technical Service Engineer Cannon Technologies, Inc.

210-222-1752 (phone) 175 CR 373 Rio Medira, TX 78066

Wheeler,Megan Conference Manager Planit

210-222-1782 (fax) [email protected] 415-455-9583 (phone)

PO BOX 2975 San Anselmo, CA 94979-2975

White,Richard

415-455-9584 (fax) [email protected] 210-353-3361 (phone)

Supervisor City Public Service, San Antonio

145 Navarro St San Antonio, TX 78205-2986

210-353-3449 (fax)

Williams,Blake

3rd Floor

210-353-3557 (phone)

Project Engineer City Public Service, San Antonio

145 Navarro St San Antonio, TX 78205-2986

210-353-4449 (fax) [email protected]

Wilson,Alan Director, Technology Management Doble Engineering Co.

617-926-4900 (phone) 85 Walnut St Watertown, MA 02472-4037

Woodcock,David 4011 Power Inn Rd

617-926-0528 (fax) [email protected]

Weidmann ACTI, Inc.

Sacramento, CA 95826-4342

Zid,Joseph Manager Exelon Corporation

630-437-2844 (phone) 2 Lincoln Ctr Oakbrook Terrace, IL 60181-4295

Zummerman,Conrad Butch General Manager CZAR, Inc.

630-576-6355 (fax) [email protected] 888-535-6261 (phone)

330 S Greenwood St Porterville, CA 93257

559-783-8998 (fax) [email protected]

End of Report

Last Achorn Alstat Andress Anglese Ashenberger Barbee Beard Benke Blake Boules Braatz Bradshaw

First John Dale Rick Rick Richard Elvis Bobby James Dennis Steve Oran Jim

Bramen Breeden Jr. Brooks Casper Clarius Jr. Corwin Cox Davis DoGraw DelaHoussaye Diete Dighe

Title

Company Florida Power & Light Co. Alliant Energy PECO Energy Alliant Energy Entergy Carolina Power & Light Muscatine Power & Water Prod. Line Support Eng. MgCutler-Hammer Pennsylvania Transformer Marketing Manager Carolina Power & Light Alliant Energy Siemens Manager Field Service

Address Address Listing 2300 Hammock Place, Sarasota, FL 34235 219 West Water, Decorah, IA 52101 344 Clearbrook Ave, Landsdowne, PA 19050 201 North Second St, Clinton, IA 52733 1801 North Falls Blvd, Wynne, AR 72396 103 Executive Parkway, New Bern, NC 28562 3205 Cedar St, PO Box 899, Muscatine, IA 52761 170 Industry Dr, Pittsburg, PA 15275 PO Box 440, 30 Curry Ave, Canonsburg, PA 15317 7600 Capital Blvd, Raleigh, NC 27804 883 West Scott St, Fond du Lac, WI 54937 PO Box 6289, Jackson, MS 39288-6289

Phone 941-331-4036 563-380-4148 610-517-9637 563-244-9656 870-261-6394 252-636-6321 563-263-2631 412-494-3027 724-873-2123 919-878-1469 920-929-6730 601-932-9835

Jeff Roberts Transformer Knoxville Utilities Board Roy Dewayne Entergy Bobby Concorde Science & Tech John CLECO Mike Richmond Power & Light TXU Electric Randy Electro-Test, Inc. Bobby Ronald Converter Station Maint. SuIntermountain Power Denis Superintendent of Metering CLECO Marc Lower Colorado River Authority Ramesh Engineer Associate II Lower Colorado River Authority

2234 S. Apapka Blvd, Apapka, FL 32703 4517 Middlebrook Pike, Knoxville, TN 37921 Grant & Barraque Streets, Pine Bluff, AR 71601 188 E. Bergen Place, Red Bank, NJ 07701 3211 Frontage Rd, Pineville, LA 71360 PO Box 908, Richmond, IN 47375 115 North Loop 12 7045 Villa Ridge Court, Grandberry, TX 76049 850 W. Brush Wellman Rd, Delta, UT 84624 304 Darnell Rd, New Ideria, LA 70560 3505 Montopolis Dr, Austin, TX 78744 3505 Montopolis Dr, Austin, TX 78744

407-880-2524 pgr: 407-941-3374 865-558-2565 870-543-5422 732-741-8989 318-641-8470 765-973-7200 972-721-6229 817-326-4028 435-864-6704 337-369-8218 572-356-6078 512-482-6342

Fabrizi Feuerhelm

Mike James

Genutis Gilbert Glidden

Donald Dick Phil

Managing Director

Golarz Gonzales Goulet Hall Hamid Harquail Hartzell

Jeff Joe Mario Melvin Joey Gregory Lyle

INCON T&D Maint. Superintendent Farmington Electric Hydro Quebec Entergy Carolina Power & Light President Concorde Science & Tech Hartzell Engineering

483 Bay St, 15th Fl. N. Tower, Toronto, ON Canada MSG2P5 416-345-5175 PO Box 817, La Crosse, WI 54602-0817 608-787-1217 909-628-1253 PO Box 2338, 4571 State St, Montclair, CA ext.231 4840 Ballard Rd, Ft. Myers, FL 33905 941-690-2705 PO Box 932, 33 State St, Bangor, ME 04402 207-941-6594 297-283-0156 PO Box 638, Saco, ME 04072 ext.217 101 Browning Prkwy, Farmington, NM 87401-7995 505-599-8333 200, rue Jean-Prouix, Hull, QC J8Z1V8, Canada 514-251-6661 5115 Thibault Rd, Little Rock, AR 72206 501-490-4708 7600 Capital Blvd, Raleigh, NC 27804 919-878-1461 188 E. Bergen Place, Red Bank, NJ 07701 732-741-8989 233 Wilhelm Rd, West Alexander PA 15376 724-663-5121

Hydro One Dairyland Power Cooperative

Harvil Heil Hizourn Hora Houck Hoyt Jr.

Terry Edgar Kacern Randy Pete Lynn

Transmission Specialist

Irnler Janicki Jared Jones Kahl

Chuck James Marc Thomas Darren

Technician

Hampton Tedder Tech. Serv. Florida Power & Light Co. Bangor Hydro

Southern Company Rohe International Inc. Entergy Alliant Energy Dow Corning Louisville Gas & Electric

224 S. Holmes Rd, Bin 76001, Rome, GA 30161 349 Northgale Dr, PO Box 2096, Warrendale, PA 15086 PO Box 2951, T-BMT-SC2, Beaumont, TX 77704 1001 Shaver Rd, N.E, Cedar Rapids, IA 52402 PO Box 995, Midland, MI 48686 4664 Jennings Lan, Louisville, KY 40218

City of Niles Utilities Dept. United Illuminating Company Western Farmers Elect. Coop. Georgia Power BC Hydro

1265 S. 15th, Niles, MI 49120 801 Bridgeport Ave, Shelton, CT 06484 PO Box 429, 701 N.E. 7th St., Anadarko, OK 73005 285 Greenville St., Newnan, GA Bag 5700, Revelstoke, BC VOE250, Canada

1

706-236-1488 pgr: 770-671-7955 cell: 770-547-2961 724-933-0260 409-635-2185 319-786-1945 517-495-4666 502-364-8675 616-684-3802 ext.307 203-920-4573 405-247-3351 770-254-7160 250-814-6625

Fax email 941-331-4030 [email protected] 563-382-6545 [email protected] 610-490-6813 870-261-6369 252-636-6328 563-262-3345 412-494-3422 724-873-2570 919-878-1485 920-929-6709 601-936-9151

[email protected] [email protected] [email protected]

[email protected]

407-884-7611 865-558-2079 [email protected] 870-543-5416 732-741-0804 318-641-8480 [email protected] 765-973-7286 [email protected] [email protected] 435-864-6618 [email protected] 337-369-8307 [email protected] 572-482-6382 512-482-6382 [email protected] 416-345-5443 [email protected] 608-787-1490 909-628-6375 [email protected] 941-690-2716 207-945-0004 [email protected] 297-283-0158 505-599-8329 514-251-7821 501-490-4740 919-878-1485 732-741-0804 724-663-5899

[email protected] [email protected] [email protected] [email protected] [email protected]

706-236-1189 [email protected] 412-933-0266 409-185-2301 [email protected]

502-364-8636 [email protected] 616-683-7964 [email protected] 203-926-4523 [email protected] 405-247-4453 770-254-7121 250-837-5624 [email protected]

Address Listing Keith Kennedy Kensmore Kimmel King Jr. Kleinert Kuchenecker Lanier Lau

Dave Patrick Steve Charles Curtis Stefan Donald Jeff Marvin

Lessard Loranz Luna Magby Marraccini McKenzie

Manager Field Projects

Roberts Transformer New York Power Authority Alliant Energy Chugach Electric Association Consolidated Edison Rohe International Inc. E-Manufacturing Carolina Power & Light LADWP

2234 S. Apapka Blvd, Apapka, FL 32703 PO Box 191, Marcy, NY 13403 935 WBR Townline Rd, Beloit, WI 53511 PO Box 196300, Anchorage, AK 99519 32 Woodland Ave, New Rochelle, NY 10805 349 Northgale Dr, PO Box 2096, Warrendale, PA 15086 7275 Industrial Park Blvd, Mentor, OH 44060 1701 W. Granlham St, Goldsboro, NC 27530 1630 N. Main St, Bldg 7, Rm 304, Los Angeles, CA 90012

Ives Al Roger Terry Ken Carl

Alstom Canada, Inc. Dairyland Power Cooperative LADWP Entergy Pennsylvania Transformer AEP

1400 Industrial Blvd, Laprairie, QC J5R2E5, Canada PO Box 817, La Crosse, WI 54602-0817 1630 N. Main St, Bldg 7, Rm 304, Los Angeles, CA 90012 5115 Thibault Rd, Little Rock, AR 72206 PO Box 440, 30 Curry Ave, Canonsburg, PA 15317 PO Box 1428, Ashland, KY 41105-1428

Mignano Miller Munroe Myers Myles

Tony Ted Mike Bill Donald

Public Service Electric & Gas 200 Boyden Ave, Testing Services, Maplewood, NY 07040 Muscatine Power & Water 3205 Cedar St, PO Box 899, Muscatine, IA 52761 Munroe Equipment Sciences, Inc. 445 Kensington St, Winnipeg, Manitoba R3J1J8, Canada Finepoint Marketing, Inc. PO Box 950, Kenwood, CA 95452 AEP 5001 Westerville Rd, Columbus, OH 43231

Nasrallah Norris

Emile Ronnie

Alstom Canada, Inc. Entergy

1400 Industrial Blvd, Laprairie, QC J5R2E5, Canada 2415 Northwest Ave, El Dorado, AR 71730

Pack Peters Picagli Pilny Plehinger Plonka Rathje Ritter Robison Rohner Rousseau Saiz Santiago Schneider Schumacker Shanks Sharp Shockites Sicker Simmermon Smith Stephens Stephenson Thomas Thornton Torres Traylor Jr. Vitiello Votimer Vonderlin

Michael Earl Tony Gerry Steve Frank Marvin Darren Jeff Bonno Serge Steve Carlos Dennis Dick John Jay Steve Robert William Cory Gary Alan Donald Robert Larry Lloyd Sal Kent Larry

Santee Cooper Richmond Power & Light United Illuminating Company New York Power Authority AEP Electrical Testing Services Alliant Energy PECO Energy Chugach Electric Association Rohe International Inc. Hydro Quebec Public Service Co. of New Mexico PECO Energy Alliant Energy Alliant Energy Entergy Knoxville Utilities Board Weidmann-ACTi Inc. First Energy First Energy Lower Colorado River Authority AEP ENMAX City of Calgary Entergy PECO Energy Western Area Power Administ. Virginia Power AEP Central Power Electric Coop. Western Area Power Administ.

1835 19th Avenue North, Myrtle Beach, SC 29577 PO Box 908, Richmond, IN 47375 801 Bridgeport Ave, Shelton, CT 06484 PO Box 191, Marcy, NY 13403 3601 Callicoate Rd, Corpus Christi, TX 78410 PO Box 7727, New Castle, PA 16107 208 South Elm St, Anamosa, IA 52205 2001 Gallagher Rd, Plymouth Meeting, PA 19462 PO Box 196300, Anchorage, AK 99519 349 Northgale Dr, PO Box 2096, Warrendale, PA 15086 200, rue Jean-Prouix, Hull, QC J8Z1V8, Canada PO Box 1268, Santa Fe, NM 87504 2001 Gallagher Rd, Plymouth Meeting, PA 19462 883 West Scott St, Fond du Lac, WI 54937 1001 Shaver Rd, N.E, Cedar Rapids, IA 52402 1801 North Falls Blvd, Wynne, AR 72396 4517 Middlebrook Pike, Knoxville, TN 37921 3580 Progress Dr. Suite K1, Bensalem, PA 19020 1910 West Market St, Bldg 4, Akron, OH 44313 1910 West Market St, Bldg 4, Akron, OH 44313 3505 Montopolis Dr, Austin, TX 78744 110 E. Wayne St, Ft. Worth, IN 46801 1200 10201 Southport Rd SW,Calgary,AB T2W4X9,Canada 5115 Thibault Rd, Little Rock, AR 72206 2001 Gallagher Rd, Plymouth Meeting, PA 19462 7940 Sorento Rd, Elverta, CA 95626 2400 Grayland Ave, Richmond, VA 23220 PO Box 1428, Ashland, KY 41105-1428 525 20th Avenue SW, Minot, ND 58701 7940 Sorento Rd, Elverta, CA 95626

President Senior Electrician

President

Station Staff Engineer

Foreman Central Zone

2

407-880-2524 pgr: 407-941-3374 315-724-8186 608-364-6547 907-762-7661 914-376-1565 724-933-0260 440-951-0900 919-583-2097 213-367-7321 450-659-8921 etx:415 608-787-1217 213-367-7324 501-490-4708 724-873-2390 606-327-1375 973-761-1810 ext:4 563-263-2631 204-779-2200 707-833-6288 614-552-1902 450-659-8921 ext:226 870-864-3880 843-448-2411 ext:37463741 765-973-7200 203-926-4702 315-724-8186 361-242-3736 724-657-5330 319-462-3380 610-941-1915 907-762-4823 724-933-0260 819-772-6042 505-473-3290 610-941-1915 920-929-6730 319-786-1945 870-261-6394 865-558-2565 215-639-8599 330-384-5685 330-384-4727 512-356-6471 219-425-2248 403-514-2621 501-490-4708 610-941-1915 916-353-4652 804-257-4661 606-327-1367 701-852-4407 916-353-4652

407-884-7811 [email protected] 608-364-6517 907-762-4694 [email protected] 914-964-8977 [email protected] 412-933-0266 440-951-0965 919-878-1485 [email protected] 213-367-7322 450-659-3371 608-787-1490 [email protected] 213-367-7322 501-490-4740 [email protected] 724-873-2570 606-327-1385 973-275-1938 [email protected] 563-262-3345 204-779-2424 [email protected] 707-833-2329 [email protected] 614-901-1820 450-659-3371 [email protected] [email protected] 843-626-6370 [email protected] 765-973-7286 203-926-4524 [email protected] [email protected] 361-242-3745 [email protected] 724-657-5332 [email protected] 319-462-3652 [email protected] 610-941-1913 907-762-4694 [email protected] 412-933-0266 819-772-6800 [email protected] 505-473-3357 610-941-1913 920-929-8709 319-786-1959 [email protected] 870-261-6369 [email protected] 865-558-2079 [email protected] 215-693-8577 [email protected] 330-384-4795 [email protected] 330-384-4795 [email protected] 512-482-6382 [email protected] 219-425-2297 [email protected] 403-514-2648 [email protected] 501-490-4740 [email protected] 610-941-1913 916-991-7114 804-257-4750 [email protected] 606-327-1385 701-852-3847 916-381-7114 [email protected]

Address Listing

Walker

Mark

Santee Cooper

1835 19th Avenue North, Myrtle Beach, SC 29577

Webber Welsh White White Williams Young Zainori Zimmerman

Jim Mike John Randy Tim Craig Adib Butch

INCON Alliant Energy AEP City of Garland Electric Western Farmers Elect. Coop. First Energy Entergy CZAR, Inc.

PO Box 638, Saco, ME 04072 1001 Shaver Rd, N.E, Cedar Rapids, IA 52402 825 Tech Center Drive, Gahanna, OH 43230 521 East Avenue B, Garland, TX 75040 PO Box 429, 701 N.E. 7th St., Anadarko, OK 73005 3601 Ridge Road, Cleveland, OH 44102 639 Loyota Ave, PO Box 61000, New Orleans, LA 70113 330 South Greenwood St, Porterville, CA 93257

Asst. Supt./Substations

3

843-448-2411 ext:3707 207-283-0155 ext:217 319-786-1945 614-552-1934 972-205-3730 405-247-3351 216-634-7364 504-576-6209 888-535-6261

843-626-6370 [email protected] 207-283-0158 319-786-1959 614-552-1939 972-205-3575 405-247-4453 216-634-7348 504-576-6029 559-783-8998

[email protected] [email protected] [email protected]

[email protected]

About EPRI EPRI creates science and technology solutions for the global energy and energy services industry. U.S. electric utilities established the Electric Power Research Institute in 1973 as a nonprofit research consortium for the benefit of utility members, their customers, and society. Now known simply as EPRI, the company provides a wide range of innovative products and services to more than 1000 energy-related organizations in 40 countries. EPRI’s multidisciplinary team of scientists and engineers draws on a worldwide network of technical and business expertise to help solve today’s toughest energy and environmental problems. EPRI. Electrify the World

© 2002 Electric Power Research Institute (EPRI), Inc. All rights reserved. Electric Power Research Institute and EPRI are registered service marks of the Electric Power Research Institute, Inc. EPRI. ELECTRIFY THE WORLD is a service mark of the Electric Power Research Institute, Inc. 1002650 Printed on recycled paper in the United States of America EPRI • 3412 Hillview Avenue, Palo Alto, California 94304 • PO Box 10412, Palo Alto, California 94303 • USA 800.313.3774 • 650.855.2121 • [email protected] • www.epri.com