662
Guidelines for partial discharge detection using conventional (IEC 60270) and unconventional methods
Working Group D1.37
August 2016
GUIDELINES FOR PARTIAL DISCHARGE DETECTION USING CONVENTIONAL (IEC 60270) AND UNCONVENTIONAL METHODS WG D1.37
Members
E. Gulski, Convenor (CH), W. Koltunowicz, Secretary (AT), T. Ariaans (NL), G. Behrmann (CH), R. Jongen (CH), F. Garnacho (ES), S. Kornhuber (DE), S. Ohtsuka (JP), F. Petzold (DE), M. Sanchez‐Uran (ES), K. Siodla (PL), S. Tenbohlen (DE)
Copyright © 2016 “All rights to this Technical Brochure are retained by CIGRE. It is strictly prohibited to reproduce or provide this publication in any form or by any means to any third party. Only CIGRE Collective Members companies are allowed to store their copy on their internal intranet or other company network provided access is restricted to their own employees. No part of this publication may be reproduced or utilized without permission from CIGRE”.
Disclaimer notice “CIGRE gives no warranty or assurance about the contents of this publication, nor does it accept any responsibility, as to the accuracy or exhaustiveness of the information. All implied warranties and conditions are excluded to the maximum extent permitted by law”.
ISBN: 978-2-85873-365-1
Guidelines for PD Detection
Guidelines for Partial Discharge Detection using Conventional (IEC 60270) and Unconventional Methods Contents LIST OF FIGURES............................................................................................................................................. 4 LIST OF TABLES............................................................................................................................................... 7 EXECUTIVE SUMMARY ................................................................................................................................... 8 1
INTRODUCTION ...................................................................................................................................... 9
2
GAS-INSULATED SWITCHGEAR (GIS) .................................................................................................... 12 2.1 Typical PD sources in GIS......................................................................................................................................... 12 2.2 Conventional electrical PD Measurement according to IEC 60270 ................................................................. 13 2.3 Unconventional PD measurement: acoustic methods ........................................................................................... 14 2.4 Unconventional PD measurement: UHF method.................................................................................................... 19 2.5 Case Studies ............................................................................................................................................................... 25 2.5.1 Case Study 1: External signals trigger alarms on UHF PD monitoring system ......................................... 25 2.5.2 Case Study 2: TOF location of PD recorded by UHF PD monitoring system ............................................ 27 2.5.3 Case Study 3: TOF location of PD signals in a cable bushing..................................................................... 29 2.5.4 Case Study 4: External EMI recorded by UHF PD monitoring system ....................................................... 30 2.6 Advantages and disadvantages of different methods ...................................................................................... 32 2.7 Summary and important aspects ............................................................................................................................ 33
3
POWER CABLE SYSTEMS ....................................................................................................................... 34 3.1 Typical PD sources in power cable systems .......................................................................................................... 34 3.2 PD pulse propagation............................................................................................................................................... 34 3.3 Conventional electrical PD measurement according to IEC 60270 ................................................................. 36 3.4 Unconventional HF PD-measurement with coupling capacitor .......................................................................... 38 3.4.1 Single-ended measurement and PD origin localization ................................................................................ 38 3.4.2 Double-ended measurement and PD origin localization .............................................................................. 40 3.5 Application of wideband HFCT for UHF PD detection at joints and terminations ........................................ 44 3.5.1 Application of HFCT at ground straps of terminations and joints ............................................................... 44 3.5.2 HFCT application at cross bonded joints ......................................................................................................... 45 3.6 System calibration with sensitivity and performance check .............................................................................. 46 3.6.1 Conventional electrical PD measurement according to IEC 60270 ............................................................ 46 3.6.2 HF PD-measurement with coupling capacitor ................................................................................................. 47 3.6.3 Sensitivity check of on-line PD measurements applied to MV grids........................................................... 55
Page 2
Guidelines for PD Detection
3.6.4 Sensitivity and performance check of unconventional PD system ............................................................... 57 3.7 Case studies ................................................................................................................................................................ 61 3.7.1 Case Study 1: Maintenance test of a 66 kV power cable with single-ended PD detection and PD localisation....................................................................................................................................................... 61 3.7.2 Case Study 2: After-laying test of a 10 kV power cable with single-ended PD detection and PD localisation ............................................................................................................................................................. 62 3.7.3 Case Study 3: Double-ended PD measurement on medium voltage cable .............................................. 63 3.7.4 Case Study 4: After-laying test 220 kV cable with single-ended PD measurements at both ends..... 64 3.7.5 Case Study 5: Continuous PD monitoring experience in a MV cable system ........................................... 65 3.7.6 Case Study 6: Comparison of insulation defect at a 66 kV GIS cable joint detected using HFCT and HF antenna .................................................................................................................................................... 67 3.7.7 Case Study 7: PD testing of short 220 kV XLPE cable system .................................................................... 69 3.8 Advantages and disadvantages of different methods ...................................................................................... 71 3.9 Summary and important aspects ............................................................................................................................ 72 4
POWER TRANSFORMERS ...................................................................................................................... 73 4.1 Types of partial discharges ..................................................................................................................................... 73 4.1.1 Pulse shape and frequency content .................................................................................................................. 73 4.2 Electrical PD measurement according to IEC 60270 .......................................................................................... 75 4.2.1 Signal attenuation ................................................................................................................................................ 77 4.3 UHF PD-measurement ............................................................................................................................................... 79 4.3.1 Signal attenuation ................................................................................................................................................ 80 4.3.2 Sensor sensitivity ................................................................................................................................................... 81 4.3.3 Recommendation for standardized valves for retrofit of UHF sensors ...................................................... 84 4.3.4 Recommendation for a dielectric window for installation of UHF sensors ................................................ 85 4.3.5 Interpretation ........................................................................................................................................................ 86 4.4 Acoustic PD-measurement......................................................................................................................................... 86 4.4.1 Signal attenuation ................................................................................................................................................ 86 4.4.2 PD localization ...................................................................................................................................................... 87 4.5 Case studies ................................................................................................................................................................ 89 4.5.1 Case Study 1: On-line PD measurement in a single-phase power transformer 220 kV/132 kV, 80 MVA.................................................................................................................................................................. 89 4.5.2 Case Study 2: Combination of conventional and UHF PD detection methods.......................................... 91 4.5.3 Case Study 3: Combination of conventional and UHF partial discharge detection methods ............... 95 4.5.4 Case Study 4: Combination of conventional method with acoustic method ........................................... 100 4.5.5 Case Study 5: Localization of PD by acoustic and UHF-measurement ................................................... 102 4.5.6 Case Study 6: Monitoring by UHF PD-measurement.................................................................................. 105 4.6 Advantages and disadvantages of different methods ................................................................................... 108 4.7 Summary and important aspects ......................................................................................................................... 108
5
REFERENCES ........................................................................................................................................ 110
Page 3
Guidelines for PD Detection
LIST OF FIGURES Figure 1: Typical PD sources in GIS along with UHF & acoustic sensors, typical interference sources ................... 12 Figure 2: IEC60270 standard test circuit [2] ...................................................................................................................... 13 Figure 3: Manual location of PD using commercially available acoustic receiver ...................................................... 15 Figure 4: piezoelectric acoustic energy sensors for picking up PD signals in GIS ....................................................... 15 Figure 5: Sensitivity of acoustic PD detection in a 300 kV GIS [5] ................................................................................ 16 Figure 6: Acoustic PD patterns; amplitude vs. time at left, PRPD on the right ............................................................. 16 Figure 7: acoustic (ultrasonic) PD pulse and enclosure reflections .................................................................................. 18 Figure 8: Narrowband vs. broadband application of the UHF PD method ................................................................. 19 Figure 9: Example equipment set-up for making on-site PD measurements using the UHF method ........................ 20 Figure 10: Typical time-domain GIS PD signal; decay time is approx. 600 ns .......................................................... 21 Figure 11: Frequency response of a simple straight section of GIS (0 – 2000 MHz)................................................ 21 Figure 12: Examples of UHF PD sensors.............................................................................................................................. 22 Figure 13: PD sensor sensitivity shown using real PD signal (MOVING particle), high-quality PRPD plot (r)........ 23 Figure 14: External UHF couplers for use at GIS spacer casting inlet .......................................................................... 23 Figure 15: Principle of time-of-flight (TOF) localization .................................................................................................. 24 Figure 16: CIGRÉ sensitivity verification: lab part 1 (l), on-site part 2 (r) ................................................................... 25 Figure 17: PD monitoring system display showing 120° offset in signals, Y-phase pattern for comparison ........ 26 Figure 18: Single-line diagram showing PD sensor locations (l), tof measurement (r)................................................ 26 Figure 19: Method for verifying TOF path delay time and direction using an externally injected pulse ............. 27 Figure 20: TOF path delay verified with injected pulse; also note the order of arrival reverses .......................... 27 Figure 21: PD monitoring display showing patters suggesting a defect in solid insulation....................................... 28 Figure 22: PRPD taken at the same sensor using UHF method (zero-span).................................................................. 28 Figure 23: Location of sensors used for TOF measurement (L); photo of sensors –Q56Y & PD30Y (R) ................. 28 Figure 24: Plot of TOF of Fig. 21: red trace is the signal from the lower sensor (PD30Y) ....................................... 29 Figure 25: UHF PD monitoring system display showing PRPD suggesting surface discharge activity ..................... 29 Figure 26: TOF set-up to locate cable bushing pd, photo of external sensor #1 (r) ................................................. 30 Figure 27: Plot of time-of-arrival of TOF diagramed in figure 26 above: approx. 38 ns ...................................... 30 Figure 28: UHF PD Monitoring system displaying signals strongly suggesting external EMI.................................... 31 Figure 29: Measurements of external EMI picked up by a UHF PD monitoring system............................................. 31 Figure 30: low-level pulsed emi interfering with cigré sensitivity verification check .................................................. 32 Figure 31: causal emi, spectrum (left), oscilloscope plot showing digital modulation (right) .................................... 32 Figure 32: Frequency-dependent phase velocity and attenuation parameters used for cable model calculation [31] ......................................................................................................................................................................... 35 Figure 33: Test circuit for measurement with a coupling capacitor at the cable termination ................................... 36 Figure 34: Measured and Simulated PD magnitude as function of travelling distance, normalized to 500 pC for the first value ............................................................................................................................................................... 37 Figure 35: Typical example of a PD pattern measured during an on-site cable test ............................................... 38 Figure 36: Measurement setup for single sided PD measurement ................................................................................. 39 Figure 37: PD source location by tdr analysis of the two travelling waves A and B with speed v ......................... 39 Figure 38: Near-end PD fault location with reflection from far end on a 500 m cable ........................................... 40 Figure 39: Setup for double-ENDED PD measurement with localization CAPABILITY ................................................ 41 Figure 40: Setup for the double-ended PD measurement with fault localization functionality ................................ 41 Figure 41: Normalized PD attenuation for double- and Single-ended measurement (10 km). ............................... 42 Figure 42: Normalized PD amplitude attenuation with a fault location at 2 km from the left end. ....................... 43 Figure 43: Double-end PD measurement data with PD origin at 220 m from Detector A and 424 m from Detector B .................................................................................................................................................................................. 43
Page 4
Guidelines for PD Detection
Figure 44: H-field at the Middle of the ‘Joint’ Section ..................................................................................................... 44 Figure 45: H-field at the End of the ‘Joint’ Section ........................................................................................................... 45 Figure 46: Model for application of hfct to a cross-bonded joint ................................................................................. 46 Figure 47: Current Level in a Cross Bonding Link .............................................................................................................. 46 Figure 48: Example of an on-site PD calibration, injecting 100 pC pulses every 10 ms .......................................... 47 Figure 49: pulse signal Measurement compared with responses calculated for different travel lengths .............. 48 Figure 50: Example of a calibration with a pulse of 200 pC on a 10,500 m long cable Unconventional PD measurement with HFCT sensors ..................................................................................................................................... 49 Figure 51: Low Frequency Response of an HFCT .............................................................................................................. 50 Figure 52: Wideband Frequency Response of an HFCT ................................................................................................. 51 Figure 53: 12 ns Square Pulse with 17 MHz HFCT (magenta) and wideband HFCT (blue) ..................................... 51 Figure 54: E-Field Coupling of Three Different HFCTs ..................................................................................................... 52 Figure 55: Input reflection of the tunnel test jig ................................................................................................................. 54 Figure 56: Tunnel test jig shown with HFCT installed for testing ..................................................................................... 54 Figure 57: Frequency Response of a well designed and shielded HFCT: green = Transfer FUNCTION; blue = return loss ..................................................................................................................................................................... 55 Figure 58: On line test to determine signal attenuation behaviour at increasing frequency ................................... 56 Figure 59: Signal attenuation versus sinusoidal frequency at transformers substations 2 and 3............................. 57 Figure 60: Setup of multi-channel PD measurement system (optical communication) ................................................. 58 Figure 61: Propagation of calibration pulses along 20 km XLPE cable line [35] ....................................................... 59 Figure 62: Propagation of calibration pulses along 20 km XLPE cable line ............................................................... 59 Figure 63: Spectra of calibration pulses ............................................................................................................................. 60 Figure 64: PD patterns of wire test ...................................................................................................................................... 60 Figure 65: PD patterns observed at 1.5x Uo during maintenance testing of a 30-year cable ............................... 61 Figure 66: PD mapping as made up to 1.5x Uo during maintenance testing of a 30-year old cable .................. 62 Figure 67: PD results of an after-laying testing of a 10 kV 2.1 km long XLPE cable section .................................. 62 Figure 68: Investigation of the joint at 955 m having PD up to 800 pC at 2xUo....................................................... 63 Figure 69: Double sided measurement on a 644 m long medium voltage cable ....................................................... 63 Figure 70: On-site testing of a 220 kV 13.3 km long XLPE cable circuit. .................................................................... 64 Figure 71: PD patterns as observed during the voltage withstand testing of a 220 kV XLPE cable ..................... 64 Figure 72: PD mapping made up to 1.3x U0 during on-site testing of a 220 kV 13.3 km long cable circuit. ......................................................................................................................................................................................... 65 Figure 73: Diagram and signals from continuous PD monitoring of a 45 kV cable system ...................................... 65 Figure 74: PD mapping of the cable system monitored................................................................................................... 66 Figure 75: frequency spectra of PD at each monitoring point of the cable system ................................................... 66 Figure 76: details of the PD source located in splice CB1 ............................................................................................... 66 Figure 77: Photograph showing PD measurement via HFCT and two HF horn antennas ........................................... 67 Figure 78: Typical simultaneous measurement result of the two HF antennas and HFCT .......................................... 68 Figure 79: Relationship between signals measured with HFCT and 2 horn antennas ................................................ 68 Figure 80: Measurement results when punched panels were not removed .................................................................. 69 Figure 81: Three-phase PRPD patterns and their equivalent 3PARD diagram ........................................................... 69 Figure 82: PRPD patterns of the PD signal at different frequencies after separation .............................................. 70 Figure 83: Findings at the GIS cable termination inspection after the failure............................................................. 70 Figure 84: Typical positive and negative PD current pulse waveforms for each degassing treatment ................. 75 Figure 85: Test circuit for measurement at a tapping of a bushing [2] ........................................................................ 76 Figure 86: Typical example for a PRPD pattern generated in a PD measuring system ........................................... 77 Figure 87: Experimental set up showing moveable PD signal source............................................................................ 78 Figure 88: Signal power dependency on location and measuring FREQUENCY ........................................................ 78 Figure 89: dependency of MEASURED APPARENT CHARGE ON LOCATION AND MEASURING FREQUENCY .............................................................................................................................................................................. 79 Page 5
Guidelines for PD Detection
Figure 90: Principle of UHF PD measurement in comparison to electrical measurement ........................................... 80 Figure 91: POSITION DEPENDENCE ON UHF SIGNAL ATTENUATION INSIDE A 210 MVA TRANSFORMER ...... 81 Figure 92: examples of uhf pd sensors for hv transformers ............................................................................................ 82 Figure 93: Typical antenna factor of a UHF PD sensor measured in GTEM cell ......................................................... 82 Figure 94: Directivity characteristic of a typical cone-shaped UHF probe simulated at 500 MHz ........................ 83 Figure 95: dependency of antenna factor on insertion depth for UHF PD sensor in oil-filled GTEM cell .............. 83 Figure 96: Potential valve types for retrofit of UHF sensors; valves with straight-through opening/duct ............. 84 Figure 97: Oil valves without straight opening/duct; retrofit of valve-type uhf sensors not possible .................... 84 Figure 98: UHF Plate Sensor with stainless steel flange and dielectric window ......................................................... 85 Figure 99: Example for dimensions of stainless steel flange and dielectric window.................................................. 85 Figure 100: Typical example of a transformer PRpD pattern obtained using uhf techniques ................................. 86 Figure 101: Illustration of the structure-borne path problem .......................................................................................... 86 Figure 102: ACOUSTIC PD SIGNAL WITH KNOWN ARRIVAL TIME HIGHLIGHTED PLUS STRUCTUREPATH component’ ..................................................................................................................................................................... 87 Figure 103: External acoustic sensors on a transformer tank with a PD inside using Cartesian coordinates ........ 87 Figure 104: comparison between pure acoustic and UHF-triggered acoustic PD acquisition [59].......................... 88 Figure 105: a) sensor PD_S1 in the 220 kV bushing, b) sensor PD_S2 in the 132 kV bushing, c) sensor PD_S3. HFCT placed at the tank earth connection ........................................................................................................... 89 Figure 106: Internal defect detected. Insulation damage due to heating ................................................................... 91 Figure 107: 15/7 MVA, 66/22 kV transformer ............................................................................................................... 92 Figure 108: long-term On-line DGA Results of 15/7 MVA, 66/22 kV transformer.................................................. 92 Figure 109: Tap sensor for PD measurements ................................................................................................................... 93 Figure 110: UHF antenna installed in oil drain valve ....................................................................................................... 93 Figure 111: Three phase PD trend of 15/7 MVA, 66/22 kV transformer .................................................................. 93 Figure 112: Three phase synchronous PRPD patterns from 15/7 MVA, 66/22 kV transformer ............................. 94 Figure 113: Equivalent 3PARD diagram ............................................................................................................................. 94 Figure 114: Individual PRPD patterns of the selected clusters ........................................................................................ 94 Figure 115: Frequency sweep diagram .............................................................................................................................. 94 Figure 116: PRPD pattern ...................................................................................................................................................... 94 Figure 117: 130/130/100 MVA – 230/115/48 kV Transformer .............................................................................. 95 Figure 118: PD tap sensor ..................................................................................................................................................... 95 Figure 119: UHF sensor placed between phases V and W ............................................................................................ 95 Figure 120 Three phase PD trend from the 130/130/100 MVA – 230/115/48 kV Transformer ...................... 96 Figure 121: Separation of PD sources using 3PARD from the 130/130/100 MVA – 230/115/48 kV Transformer ............................................................................................................................................................................... 96 Figure 122: On-line frequency sweep diagram ................................................................................................................ 97 Figure 123: On-line frequency sweep diagram ................................................................................................................ 97 Figure 124: Trend diagram and PRPD patterns of the signal detected in the UHF range ....................................... 97 Figure 125: PD sources separation combining the conventional and unconventional measurements ...................... 98 Figure 126: Increase of the amplitude of the PD pulses detected at the phases V and W ..................................... 98 Figure 127: Acoustic PD localization .................................................................................................................................... 99 Figure 128: PD source localization in the vicinity of the phase V .................................................................................. 99 Figure 129: Dismantled bushing of phase V ...................................................................................................................... 99 Figure 130: Endoscopic inspection ........................................................................................................................................ 99 Figure 131: PD tracks at phase V ..................................................................................................................................... 100 Figure 132: PD tracks at phase W ................................................................................................................................... 100 Figure 133: MVA 150/20.8 kV power transformer and PD decoupling at the bushing tap................................ 101 Figure 134: Conventional PD measurements at U= 85 kV in phase V (upper pattern) and in phase W (lower pattern) ...................................................................................................................................................................... 101 Figure 135: Acoustic measurements and defects location ............................................................................................ 102 Page 6
Guidelines for PD Detection
Figure 136: 333 MVA transformer showing positions of UHF sensors and acoustic sensors [67] ......................... 103 Figure 137: Measured propagation time differences between three UHF probes ................................................. 104 Figure 138: Deteriorated paper insulation on leads at the tap changer.................................................................. 105 Figure 139: UHF PRPD patterns 1 – 3 [69] .................................................................................................................... 106 Figure 140: Number of PD per minute and correlation coefficient [69].................................................................... 107 Figure 141: results of recognition algorithm of determined patterns 1 - 3 [69] .................................................... 107
LIST OF TABLES TABLE 1: EVALUATION OF DIFFERENT PD MEASURING TECHNIQUES ........................................................................ 11 Table 2: Acoustic propagation velocity for different materials in GIS ......................................................................... 18 Table 3: Comparison of UHF and conventional (IEC 60270) methods (GIS) ............................................................... 32 Table 4: Typical Dimensions of Joints .................................................................................................................................. 44 Table 5: dB to Rt conversion ................................................................................................................................................... 55 Table 6: Comparison of conventional (IEC 60270) and unconventional PD detection methods for cables ........... 71 Table 7: Raw PD patterns acquired by means of three sensors (row 1); PD patterns after clustering processing (rows 2, 3, 4) ........................................................................................................................................................ 90 Table 8: Fault gases concentrations from lab tests ........................................................................................................... 98 Table 9: Comparison of UHF and conventional IEC 60270 measurement method .................................................. 108
Page 7
Guidelines for PD Detection
EXECUTIVE SUMMARY Detection and evaluation of partial discharges (PD) belong to fundamental measurements applied to HV components. Due to on-going development of their application in both laboratory and field conditions, there is a continuing need to support this process with guidelines for PD detection and testing, describing new and established methods. As a result, PD measurement has become a worldwide-accepted method for insulation diagnosis and a required part of the acceptance testing for most HV assets. Based on the absence or presence of PD activity caused by insulation defects discharging during routine tests, on-site tests, or periodic in-service inspections throughout the service life-time, conclusions may be made about the actual condition of the dielectric insulation system. Following the studies published in 2010 - CIGRÉ Technical Brochure 444 - this new publication continues the discussion on the application of conventional and unconventional partial discharge detection methods. In particular, the individual chapters of this brochure discuss several aspects of applying PD detection to different types of HV components:
Evaluation of quantities to correlate conventional (IEC 60270) PD detection to unconventional methods: a) definition of quantities and procedures to consistently correlate standardized PD [pC] and unconventional (HF) instrument reading(s), b) overview (case studies) of best-practice methods to apply and evaluate PD measurements for testing purposes of different components, c) effects of advanced noise suppression and signal processing techniques on the reading(s) sensitivity, Methods to determine the sensor sensitivity: a) evaluation procedures of parameters to describe the sensors’ sensitivity, b) frequency spectrum (magnitude, power spectrum), c) impedance of sensors versus frequency / effective height, etc.
are evaluated by means of practical examples (case studies) in both laboratory and field. In chapter 1, the general aspects and background information of conventional and unconventional PD detection as applied to different power components are presented. In chapter 2, PD defects typical for GIS and their PD pulse characteristics are evaluated for conventional and unconventional detection methods. In chapter 3, specific aspects of using different PD detection methods as well as interpretation techniques for power cables are described. In the case of unconventional detection methods, solutions are shown to demonstrate the importance of sensitivity and performance checks. In chapter 4, PD detection as used for power transformers is discussed in the scope of PD pulse shape, signal attenuation, the sensitivity of using UHF sensors. In addition, a small section covering acoustic PD detection has been included. Finally, an extensive reference list covering the most recent overview of international publications is included.
Page 8
Guidelines for PD Detection
1 INTRODUCTION PD measurements are recognized worldwide for insulation diagnosis and are usually a required part of the acceptance testing procedure for most HV equipment assets. Based on the absence or presence of PD activity caused by insulation defects discharging during routine tests, on-site tests, or periodic in-service inspections throughout the service life-time, conclusions may be drawn about the actual condition of the dielectric insulation system. To use one example, on-site PD tests can be applied to prove two main characteristics of a system of HV components: 1. HV system component quality and integrity of the system components: As part of commissioning on-site: a check for possible damage after completed factory tests due to transportation, storage and installation. To assure that no critical defects in the insulation system were caused during transport from the manufacturer to the site and during the erection on-site. Typically, the system components are tested in the factory including the main parts and any prefabricated accessories. However, the effect of transportation and the correctness of the final assembling can only be tested after completing the installation in the field. After on-site repair: to spot mistakes in workmanship and to demonstrate that the equipment has been successfully repaired and that all dangerous defects in the insulation system have been eliminated. 2. Availability /reliability of the HV component: For diagnostic purposes: to estimate actual condition of the service aged component by checking the insulation fingerprint in order to note any insulation system degradation after a period of service operation e.g. 40 or 50 years. By providing reference value (‘finger print’) for diagnostic tools (voltage test including partial discharge) for later tests to demonstrate whether the insulation system is still free from dangerous defects and that the lifetime expectation is sufficient high. To improve the sensitivity of PD measurements, different techniques/methods have been developed and they are in use for both laboratory and on-site applications. All these methods can be divided into conventional and unconventional PD measuring methods. History shows that PD measurement technology has been proven as an excellent method for quality control of HV insulation for over 40 years. Due to on-going aging of the insulation systems of HV components in service, on-site PD testing and diagnostic methods have attracted increased interest for application in condition monitoring. Since conventional PD measuring systems used in the controlled factory environment are not usually suitable for on-site application, specialized PD detection and measurement methods have been introduced. PD signals can be detected by so-called unconventional PD measuring methods and systems which utilize different physical characteristics and properties of the PD processes. In general, electric methods are based on the measurements of electrical signals in the radio frequencies (RF) ranges e.g. HF, VHF, UHF, etc. Many technical studies and papers have demonstrated that there is virtually NO correlation between the apparent charge [pC] values recorded from standardized conventional measurements and the values recorded by unconventional measurements. This is especially true when measuring with the RF method. Measurements using RF techniques are based on the detection of electromagnetic waves emitted by the PD event. There is no correlation to standardized PD measurements because the PD level recorded depends on the type of PD, the location, the sensor type and the size and geometry of the physical object in which the electromagnetic wave is measured. In particular, the measured values of a PD signal strongly depend on the insulation defect's type and geometry and on the location of the sensor in relation to the PD source.
Page 9
Guidelines for PD Detection
Since it is impossible to measure a PD event directly, remote sensors will always record values that are influenced by the fact that PD pulse propagates over some distance from its point of origin. The following effects are normal:
The increasing impact of pulse propagation path on the received PD waveform e.g. reduced rise time and increased pulse width, attenuation, etc. The superposition between different signals, e.g. multiple PD sources, crosstalk, external interferences.
Due to the fact that there is a general interest in the application of these “nonstandard” methods because of their advantages for use in the field, this guide has been prepared to provide recommendations for implementing them. This guide discusses the current state of the art in the field of PD measurements with unconventional methods. Included in the following guide, the reader will find actual results obtained for different types of HV components, with 'best practice' solutions being discussed and explained. There are several methods for generating high voltage ac (HVAC) which can be applied for on-site insulation withstand testing and for high voltage testing in combination with PD measurements, see TABLE 1. Therefore, combining PD detection with the standard on-site HVAC test provides much more comprehensive information about the state of the insulation system, especially the possibility to discover and locate defects within the insulation system. It is also known that there is no general relationship between the PD intensity and the breakdown probability. Therefore PD measurement is an indirect method which delivers an indication of the extent of degradation and the potential danger posed by the weaknesses identified. With conventional PD measurement (IEC 60270), the apparent charge is measured in pC. It is the integrated current pulse, caused by a PD, which flows through the test circuit. The conventional method allows a precise calibration, but requires a sufficiently high signal-to-noise ratio (SNR) in the measurement circuit to easily resolve the PD signal in question. The standardized method of IEC 60270 and HVAC test protocol is well established for factory and laboratory testing, but is often not appropriate for on-site testing. In the case of field-testing where very high background noise levels are present, unconventional methods which are capable of high SNR have been proven to be the best way to obtain meaningful PD measurements. Several unconventional PD detection methods based on acoustic and electromagnetic phenomenon have been used for some time for PD detection on power cables, transformers, GIS, and generators. Up until now, there have not been accepted procedures and guidelines for these 'unconventional´ methods as has been the case for conventional methods for many decades. There are many open questions including: calibration or sensitivity verification procedures, techniques for noise suppression, and methods of fault location. The authors of this guide believe that now is the time to prepare guidelines and international recommendations for these unconventional PD detection methods in order to ensure reproducible and comparable PD measurements on high voltage equipment between users throughout the industry.
Page 10
Guidelines for PD Detection TABLE 1: EVALUATION OF DIFFERENT PD MEASURING TECHNIQUES Conventional PD measurements (IEC 60270)
Unconventional PD measurements
Advantages
Disadvantages
Advantages
Disadvantages
1. Calibrated readings of apparent charge value in [pC] 2. Well-known method since 1960s 3. Reference PRPD patterns are widely available for different discharge defects and for different components 4. Standardized procedures for type-, factory- and on-site tests
1. EMI (electromagnetic interference) makes on-site application difficult 2. Size and complexity of the object influence the sensitivity of the measurements. 3. Localization of the PD source not possible 4. In the case of large (distributed) test object there is strong influence of defect position on measured PD value 5. Reduced sensitivity in the case that the coupling capacitor capacitance is low compared to that of test object
1. Better rejection of EMI on-site 2. Distributed measurements and synchronous PRPD pattern evaluation possible 3. In most cases applicable for on-line monitoring 4. PRPD patterns very similar to those obtained with conventional measurements
1. Calibrated readings of apparent charge value in [pC] not possible 2. Standardized testing procedure not available for all components 3. Strong influence of defect position on measured PD value, in particular for measurements in the VHF and UHF frequency ranges 4. Minimum number of sensors necessary to fulfil the CIGRE sensitivity recommendation
Transformers: a) Apparent charge levels not sufficient for diagnostic purposes b) Difficult application for transformer bushing without measuring tap
HF measurement for cables (use of coupling capacitors and HF coupling devices): a) Applicable off-line b) Localization of PD source possible on complete power cables using Time Domain Reflectometry (TDR)
HF measurement for cables (use of coupling capacitors and HF coupling devices): a) Not applicable on-line b) PD measurement in long cable requires measurements at both cable ends
Power Cables: a) EM interference dependent on type of HV source used on-site b) PD measurement in long cable requires measurements at both cable ends GIS: Sufficient on-site PD sensitivity fulfilled only with encapsulated voltage test set-ups (typically for GIS of lower voltage ratings)
HF/VHF measurement (use of HFCTs): a) Applicable on-line and off-line b) Sensors can be installed during operation c) PD localization possible on cable accessories
UHF measurement (Transformers, GIS): a) High signal-to-noise ratio possible b) Widely accepted for GIS testing & monitoring c) UHF sensor can be installed online for transformers Acoustic measurement (Transformers, GIS): a) Applicable on-line and off-line b) Used for PD source localization on transformers On GIS used for PD source localization and PD defect recognition; no restriction on sensor positioning
Page 11
HF/VHF measurement (use of HFCTs): a) Sensitivity and possibility of localization depends on measurement frequency selection b) Sensor installation not always possible c) PD localization in cable accessories only possible if the sensor can be installed d) Type and position of the sensor influences the sensitivity e) Comparison of measurements with different detection systems not possible UHF measurement (Transformers, GIS): a) Standardized procedure for acceptance test and sensitivity check available for GIS only b) Pre-installed internal sensors are common practice for GIS c) External sensors can sometimes be applied for GIS, but have limited sensitivity to PD and higher susceptibility to external EMI d) Signal processing necessary for noise suppression Acoustic measurement (Transformers, GIS): a) Limited sensitivity (except on GIS)
Guidelines for PD Detection
2 GAS-INSULATED SWITCHGEAR (GIS) 2.1
Typical PD sources in GIS
Figure 1 contains a pictorial summary of typical PD sources and insulation defect types which may occur in GIS and produce radio-frequency (VHF and UHF) signals.
FIGURE 1: TYPICAL PD SOURCES IN GIS ALONG WITH UHF & ACOUSTIC SENSORS, TYPICAL INTERFERENCE SOURCES
Similarly to the case with other complex insulating systems, two requirements must be fulfilled for PD to occur [1]:
The electric field strength E must exceed the withstand level of the insulation material in the immediate vicinity of the defect. Free electrons must be available to initiate and sustain the PD (the processes governing this availability are extremely complex and lead to the stochastic, non-regular behavior of PD signals).
As shown in Figure 1, typical PD sources in GIS consist of the following main types:
Moving or ‘hopping’ particles, typically metallic and in the millimeter range, e.g. a tiny fragment of aluminum (from the enclosure) chiseled off by screw-threading during assembly. It is well-known that such particles move under the influence of the electric field. The signals they generate are relatively easy to detect by both conventional electrical and unconventional (acoustic & UHF) methods. This relative ease of detectability is good because moving particles are considered a relatively high risk because of their unpredictability in GIS [2], [3]. Floating potential discharges result when a conducting (metallic) object, e.g. a field electrode or contact component, has no galvanic connection to one of the HV poles - the inner conductor or the enclosure. This forms a capacitive voltage divider which charges up under the influence of the electric field. Since the gap across the missing contact is usually very small (<<1 mm), it sparks over each time the potential difference across the gap exceeds the withstand strength of the insulation (e.g. SF6), thus momentarily short-circuiting the floating component, but which then immediately begins charging up again. Floating potential PD usually presents high amplitude and high pulse count signals. Protrusions, on either the HV inner conductor or on the enclosure of the GIS, produce PD signals with the well-known characteristics of corona discharge. Protrusions are normally found during factory testing and thus turn up relatively rarely on site.
Page 12
Guidelines for PD Detection
Defects in solid insulation components, e.g. the insulators which support the GIS inner conductor, insulated drive shafts, bushings, cable end connections, etc., typically consist of single, isolated voids or they take the form of cracks or delamination near the interface between insulation and metal parts. Particularly in the case of voids, lack of a start electron to initiate PD often leads to long delays in PD inception time; this delay effect may prevent them from being detected during the usual factory HV test intervals (typically 60 s). Small isolated voids (<1 mm) typically generate PD signals of very low amplitude (<<10 pC) but which can also exhibit high variation in their activity over time. Cracks or delamination may also produce PD signals, but in the case they become filled with SF6 gas, this may make such defects even more difficult to detect with any method. Surface discharge PD, caused by surface contamination of insulating components, normally understood to be conductive (metallic) or semi-conductive pollutants. Surface discharge can also occur within the internal boundary layers of graded-potential bushing structures, where the GIS is connected to cables and transformers. Surface discharge PD usually presents relatively high signal magnitudes, but the physics of surface discharges are also quite complex and thus the signals may exhibit high variability over time. A special case of surface PD is when a small isolated metallic particle is laying on an insulator surface. Such defects produce PD signals of very low magnitude and are thus very difficult to detect.
PD pulses in SF6 exhibit very fast rise-times of less than 100 ps [4] and fall-times (i.e. the ‘tail’ of the pulse) of one nanosecond or so. The short rise-time is due to the initial rapid avalanche process, while the slower fall-time is due to the slower space charge drift velocity of electrons and ions in the gas [4], [5]. Viewed in the frequency domain, these fast rise-times transform into RF (radio frequency) signals whose spectral components extend well past the UHF band (300 – 3000 MHz). For more than 30 years, these techniques have been developed and refined to detect PD activity within GIS using these RF methods [6], [7], [8], [9], [10], [3], [11]. The RF signals we actually pick up in GIS do not at all resemble the clean, elegant PD pulse shapes predicted by theory or reported on by Judd and others [12], [7], [8], [9], even when the RF PD sensor is positioned adjacent to the PD source. This is because the fast rise-time PD pulses immediately propagate away from the PD source site and throughout the interior of the equipment, exciting a multiplicity of constructive and destructive resonances which can take up to several hundred nanoseconds to decay. The original RF pulse shape is already heavily distorted by these resonances immediately after its emission. Further explanation concerning propagation effects follows below.
2.2
Conventional electrical PD Measurement according to IEC 60270
As with most HV equipment, factory testing of GIS is typically carried out by connecting individual shipping units (GIS modules) within the standard conventional test circuit defined in IEC60270 [13], a schematic diagram of which is shown in Figure 2 for reference. This normally works very well because an individual GIS module is a very good approximation of a lumped capacitance. PD occurring within the GIS will result in small currents flowing through the coupling capacitor and through the measuring impedance, thus being registered by the PD measurement equipment.
FIGURE 2: IEC60270 STANDARD TEST CIRCUIT [2]
Page 13
Guidelines for PD Detection
[NOTE: ‘OL optical link’ is a mistake in the original] The major advantage of this ‘conventional’ method is that IEC 60270 describes a calibration method in detail. This allows comparison of PD levels (signal amplitude) in equipment under test with set limits, thus establishing pass/fail criteria. For this reason it has become universally accepted as the standard for PD acceptance testing, and this is what is meant by ‘conventional’ PD measurement technique. For further information the reader should consult the standard itself [13], [2], [3], [11] and the open literature. The principle disadvantages of IEC 60270 are as follows:
The maximum measurement frequency specified by IEC 60270 is 1 MHz ([2], 2014 revision) which translates to a wavelength of 300 m, thus making localization of PD sources impossible. (For this reason, manual acoustic methods are often employed during factory testing to locate PD sources.) This is of particular significance for on-site testing [5], [14], [15]. Unfortunately, coupling mechanisms for both conducted and radiated EMI function very efficiently in the usual set-ups used for conventional PD measurement. Typical examples are external corona from the HV source itself, medium-wave (AM) broadcast signals which couple into the test circuit connections (which often form efficient antennas and electrical noise on power supply and ground connections. Such background interference signals often exceed the acceptance level for PD activity. Ridding the test setup of their effects is often technically challenging and expensive to achieve, and in some cases impossible. Lastly, and of particular significance for on-site testing of GIS, the ratio of the capacitance of the coupling capacitor capacitance to that of the test object should be as large as practicable for good sensitivity. However, applying that rule to a large-scale transmission GIS would mean a very large and heavy coupling capacitor which in turn also significantly increases the load seen by the HV power supply (this is even pointed out in IEC 60270 [13]).
These factors motivated early researchers to begin exploring the viability of alternative, 'unconventional’ techniques, such as the UHF method and acoustics.
2.3
Unconventional PD measurement: acoustic methods
Partial discharges may be thought of ‘micro-sparking’. Analogous to the way terrestrial lightning produces thunder, an acoustic shockwave is produced as the gas rushes back into the evacuated partial discharge channel which propagates away from the PD site throughout the surrounding GIS structure. Also, as is intuitively obvious, moving (hopping) particles excite characteristic acoustic transients as they impact the enclosure. These acoustic signals can be picked up using sensors active in the ultrasonic region (~10 – 500 kHz) which are placed in contact with the outside of the GIS. Most of the fundamental work establishing the methods and effectiveness of acoustic techniques for GIS PD diagnostics was done in the 1990s [16], [17], [18], [19], [15], [10], [3] and continues to the present day [20]. Acoustic methods are sensitive enough to detect PD sources such as moving particles, floating potential discharges, and protrusions usually (if they generate signals which are large enough to be detected). However, PD sources in solid dielectrics such as voids or weak surface discharges are usually quite difficult to detect with acoustic techniques because these materials exhibit high levels of attenuation to acoustic signals. In any case, high levels of pre-amplification are usually required to pick up the signals because the mechanical vibrations produced by PD are relatively weak. Acoustic methods employ two types of equipment: self-contained manual devices and ‘instrument-based’ systems.
Page 14
Guidelines for PD Detection
SELF-CONTAINED ACOUSTIC DEVICES FOR MANUAL LOCATION OF PD SOURCES An example of a ‘self-contained’ instrument for manual location of PD sources is shown in Figure 3. Operation is intuitively straightforward: the operator moves along the section of GIS under investigation by carefully pressing the pointed tip of the sensor wand against the GIS enclosure as shown in the photograph and listening for signals in the headphones. Settings on the instrument can be used to vary the way the signal is processed to enhance detectability. With some practice, an operator can discern even quite low-level PD and locate it accurately, owing to the strong dependency of acoustic signals on distance to the source; the characteristic ticking sound of even very small moving particles (<1 mm in length) can be clearly located with this method. More background can be found e.g. in [17], [18].
FIGURE 3: MANUAL LOCATION OF PD USING COMMERCIALLY AVAILABLE ACOUSTIC RECEIVER
‘INSTRUMENT-BASED’ ACOUSTIC PD SYSTEMS The second and more complex category of systems for acoustic detection of PD are typically based on using piezoelectric sensors affixed to the GIS enclosure which are connected to a specialized, purpose-built instrument. These instruments usually feature an array of sophisticated signal-processing and display algorithms to aid in assessing the signals, and they also supply power to the signal pre-amplifiers typically built into the sensors. Piezoelectric sensors used for picking up acoustic PD signals typically work in the frequency range from 10 kHz to several 100s kHz (max. 1 MHz) and consist of the following types:
Acoustic emission piezo-electric sensors. Accelerometers. Condenser microphones.
Some typical piezoelectric sensors are shown in Figure 4. Signal output is via the BNC connectors (facing left).
FIGURE 4: PIEZOELECTRIC ACOUSTIC ENERGY SENSORS FOR PICKING UP PD SIGNALS IN GIS
The SNR of acoustic PD measurements depends on the type of sensor used, the signal processing employed, and as already mentioned, the acoustic signal propagation path between the PD source and the sensor. The Page 15
Guidelines for PD Detection
acoustic signal levels from PD sources in GIS are highly dependent on the type and location of the PD source and the transmission path between that source and the sensor(s). It is often very difficult or impossible to pick up signals from a weak PD source even just one compartment away. Consequently it may be necessary to affix acoustic sensors to more than one location for large compartments, e.g. circuit breaker housings or GIL. For this reason, acoustic techniques are almost never employed for on-line monitoring, but again it makes them very useful for locating PD sources. It is generally assumed that measuring PD using acoustic methods is immune to electromagnetic interference (EMI) noise in the substation. Indeed this is one of the primary advantages of acoustic techniques and is usually valid, but operators must remember that acoustic sensors will pick up mechanical vibrations present and that these will often be synchronized with the fundamental and harmonics of the 50/60 Hz power frequency. Care must be taken to not mistake such signals for actual PD. In Figure 5 it is assumed that such a linear relationship only applies within the confines of a single GIS compartment.
FIGURE 5: SENSITIVITY OF ACOUSTIC PD DETECTION IN A 300 kV GIS [5]
While acoustic methods are typically focused on localization of PD sources, if the SNR is sufficiently high, acoustic techniques can be used to determine the PD type. As was the case with electrical methods, acoustic signals were originally amplified and simply displayed on an oscilloscope for analysis. However, more sophisticated signal processing and display algorithms are being applied to visualize the PD signals for analysis. Two of the best known are shown in Figure 6. The left-hand plot shows the algorithm developed by Lundgaard [18], [19] the so-called ‘time-of-flight’ plot (not the same as the TOF method for locating PD sources) which plots the amplitude of the acoustic signal from the impact of a moving particle vs. time. This allows an estimate of the amount of time the particle is ‘in flight’, and therefore its relative size and the risk it presents. The right-hand plot shows the acoustic PRPD pattern of a moving particle.
FIGURE 6: ACOUSTIC PD PATTERNS; AMPLITUDE VS. TIME AT LEFT, PRPD ON THE RIGHT
Page 16
Guidelines for PD Detection
It should be noted that extra care must be taken to prevent pulse pile-up, since the acoustic signals are by definition slower than e.g. RF or UHF signals. This means it is possible that successive pulses may overlap for PD sources with very high pulse counts (e.g. a floating electrode). It does not rule out PD pattern assessment with acoustic methods, but such factors must be kept in mind. Again, because of the dependency on propagation path and low sensitivity to defects in solid insulators, it is usually necessary to carefully check compartment by compartment in order to verify the presence of the weak PD signals (or their absence). Such measurements can be quite time-consuming, but in GIS with few or no RF PD sensors installed or e.g. long feeder ducts, acoustic methods may be the only possibility to carry out on-site PD assessment. ACOUSTIC TRANSMISSION PATH CHARACTERISTICS OF GIS:
Propagation of the acoustic waves between the PD source and the acoustic sensor is strongly influenced by the geometry of the test object and the materials through which the acoustic signals pass. At first glance, the metal enclosure of the GIS (typically aluminium) would appear to present a very good, lowloss transmission medium for acoustic signals PD signals, and this holds true for e.g. picking up the signal from a moving particle within the confines of a single compartment, where attenuation of the acoustic signals is low due to the stiffness of the material. However, drastic differences in elasticity, attenuation, and velocity of propagation of the various materials present in GIS result in a high proportion of the signals’ energy being reflected at compartment boundaries (flanges) and poor transmission across them. Also, acoustic signals are strongly attenuated by solid insulation materials and the SF6 gas itself. As a result of these effects, acoustic signal propagation is very complex, and acoustic PD signals tend to fall off more rapidly in amplitude than the corresponding RF (UHF) signals, especially as we move away from the PD site. This is why acoustic methods are often employed to locate PD sources during conventional (IEC 60270) factory testing of GIS. Table 2 gives some sample values for the velocity of propagation in some of the typical materials found in GIS, and Figure 7 is a plot of an acoustic PD signal clearly showing the strong reflections from the compartment boundaries.
Page 17
Guidelines for PD Detection TABLE 2: ACOUSTIC PROPAGATION VELOCITY FOR DIFFERENT MATERIALS IN GIS
Material:
Propagation velocity:
Attenuation:
Air
343 m/s
Mild
SF6
136 m/s
Strongly damping
Aluminum
6420 m/s
Very low
Epoxy
1230 m/s
Strongly damping
FIGURE 7: ACOUSTIC (ULTRASONIC) PD PULSE AND ENCLOSURE REFLECTIONS
Owing to the factors described above, it is impossible to quantitatively correlate the received acoustic signal amplitude to PD charge, even more so than is the case with RF techniques. As a result of this and similarly to the case with the UHF technique, the original ‘CIGRÉ Sensitivity Verification Method’ [21] also established a comparison procedure to verify the relative sensitivity of acoustic PD measurements. Note that it was decided to drop the acoustic check from the current revision of [21] because it was deemed too difficult to achieve a realistic level of reproducibility and this technique is not used for continuous monitoring. LOCATION OF PD SOURCES USING ACOUSTIC METHODS As already mentioned above, the main application of acoustic methods in PD measurement is for localization of the PD source. There are numerous publications detailing these methods for many types of high voltage equipment (GIS, transformers, cables, overhead line insulators, bushings, etc.) [5], [14].
In general, two different methods can be used for locating PD sources acoustically. One is to simply contact the GIS enclosure with an acoustic sensor along its length and estimate the location of the PD source by observing the changes in amplitude and high frequency content of the received signal as a function of the location of the sensor. The other method is to place two or more sensors at different locations and calculate the location of the PD source based on the sensor positions and the time-of-arrival of the acoustic signals, taking into account as well as possible the variation in acoustic signal propagation velocity along the signal paths. This usually known as the ‘time-of-flight’ (TOF) method and is essentially the same as used with RF/UHF techniques. However, the variation in propagation path characteristics for acoustic signals along their transmission paths is even higher than when using RF techniques which make acoustic TOF measurements trickier to perform outside of a single compartment. (This is why it is typically more expedient to locate the PD source by simply searching for the highest signal level, and this is often sufficient.) However, for a long section of GIL or even a circuit breaker housing, a TOF measurement may be very effective; another trick to more precisely verify a location is by gently tapping on the GIS enclosure with e.g. a small metallic object, i.e. ‘inject’ a test signal for the acoustic TOF measurement. Such tricks can help in making acoustic TOF measurements across chamber boundaries
Page 18
Guidelines for PD Detection
(flanges), which can be quite challenging because of the discontinuity in material properties and thus propagation velocity. Such ‘time-of-arrival’ methods utilizing several sensors and sophisticated localization algorithms have been developed for use on HV transformers and are described in the later chapter.
2.4
Unconventional PD measurement: UHF method
Applications of radio frequency (RF) methods to the detection, location, and monitoring of PD in GIS – usually referred to (somewhat incorrectly) as the ‘UHF method’ - began to appear in the early 1980s [7], [8], [9]. Basically these methods utilize some kind of RF sensor to pick up the RF signals produced by the PD event, amplify them and display them, often using the same algorithms (e.g. PRPD) used in conventional methods. Although many refinements and improvements have been made during the past 30 years by many workers in the field, the basics of the method have remained more or less the same. It is covered in detail below. DESCRIPTION OF THE UHF METHOD: The UHF method generally refers to techniques which pick up the fast rise-time RF signals produced by PD using some kind of near-field sensor at UHF frequencies (defined as 300 MHz to 3000 MHz). [Arguably it should be known as the ‘VHF/UHF method’ since PD signals are often picked up below the actual UHF band, but this is the term which has come to be commonly used to refer to the method.]
The sensitivity of a UHF measurement system is dependent on the following factors (from [2], [5], [14], [3], [11]):
The type of defect and its location.
GIS size and configuration, which determines the signal propagation characteristic.
Sensor location, sensitivity, and frequency response.
The measurement chain characteristics (gain/loss, frequency response, etc.).
The characteristics of the measurement instruments themselves.
UHF TECHNIQUE: NARROW-BAND VS. BROADBAND The UHF method is usually carried out as shown in Figure 8; narrow-band techniques employing variable-width frequency windows <<1% BW (e.g. 3-5-10 MHz) or wide-band techniques in which the system front-end detects and/or down-converts over wide frequency ranges (e.g. 200 MHz to 2000 MHz). The latter are typical for UHF PD monitoring systems. Some authors have also described a hybrid combination of the two which has shown certain advantages [22].
FIGURE 8: NARROWBAND VS. BROADBAND APPLICATION OF THE UHF PD METHOD
A schematic drawing of a typical narrow-band UHF PD measurement set-up for on-site diagnostics is shown in Figure 9.
Page 19
Guidelines for PD Detection
FIGURE 9: EXAMPLE EQUIPMENT SET-UP FOR MAKING ON-SITE PD MEASUREMENTS USING THE UHF METHOD
EQUIPMENT FOR PD MEASUREMENTS USING UHF TECHNIQUE Figure 9 shows some of the equipment typically used for performing PD measurements on GIS using the UHF technique. Since the PD signals are weak, RF pre-amplifiers are used to boost the low-amplitude signals and also to overcome the loss inherent even in high-quality, low-loss RF coaxial cables. Power-frequency voltages and switching transients present at the sensor output (not to mention a flashover) can potentially destroy the pre-amps, so some form of protective circuitry is typically used ahead of them. In the drawing the term ‘MUX’ refers to a multi-channel (2 or more) multiplexer which allows switching between several sensors and which is also typically used to power the pre-amps remotely. The RF pre-amps should cover the frequency range of interest (e.g. 200 MHz to 2000 MHz), generate relatively high gain (> 30 dB) and introduce as little noise as possible (noise figure < 6 dB or so).
Early workers in the field used the fastest available oscilloscopes or RF spectrum analysers to observe the PD signals in the VHF/UHF region [20], [7]. This practice continues today, especially for on-site diagnostics (and of course in the research laboratory) because the RF spectrum analyser allows observing signals across the frequency band of interest. With experience the operator can learn to discriminate between EMI and actual PD signals and even between types of PD. Although widely used, the typical swept-frequency super-heterodyne RF spectrum analyser laboratory instrument is unfortunately not ideal for observing PD signals because the latter are very broadband pulsed signals, and they are highly stochastic. In other words they resemble broad-band noise. Spectrum analysers are rather more complex instruments to use than an oscilloscope. The settings are closely interrelated (indeed they’re usually interlocked in ‘auto’ mode) and it is quite easy to make fundamental errors when measuring ‘difficult’ signals like PD. A thorough understanding of their functionality is required to use them correctly, but this is outside the scope of this chapter. (The reader is strongly encouraged to refer to [23] for more detailed background information about their operation.) Once a PD signal is identified, the spectrum analyser itself can be employed for some limited analysis. Some simple examples: wide-band spectral filling typically means the PD source is very close to the sensor, perhaps even within the same enclosure, while rapid roll-off of the spectrum suggests the source is further away, or even external to the GIS. Although possible to identify some types of PD using only the spectrum analyser, it is better to find a region where the PD signal SNR is high, switch the analyser to ‘zero span’ and use it as a tuned receiver to down-convert the signal for display on a purpose-built PD instrument, i.e. to produce the well-known PRPD pattern.
Page 20
Guidelines for PD Detection
UHF TRANSMISSION PATH CHARACTERISTIC OF GIS At first glance, GIS appears to resemble coaxial cable or waveguide, so that it could be assumed it presents a low-loss, well-behaved transmission medium for VHF & UHF signals. Unfortunately this is anything but the case. GIS is designed for transmission of electrical energy at high currents and voltages, not VHF and UHF signals. All of the internal components are carefully and painstakingly designed for safe, reliable, long-term high voltage and high current operation. However, the result is that the broadband RF pulses produced by PD encounter an extremely complex RF propagation environment filled with conductive surfaces, dielectric boundaries, and abrupt impedance mismatches and discontinuities (i.e. at bends or changes in enclosure diameter). Furthermore, the dimensions of the GIS interior are of the same order as the wavelengths of the RF signals being measured. Therefore those signals undergo profound distortion - constructive and destructive reflection, attenuation, dispersion, and leakage – as they travel along through a GIS. Complex mixtures of higher-order modes are excited by these interactions [24], [25]. Because of these effects, GIS has been characterized as ‘heavily overmoded waveguide’ [22]. The original PD pulse shape is lost, replaced by a slowly decaying complex high frequency transient as shown in Figure 10.
FIGURE 10: TYPICAL TIME-DOMAIN GIS PD SIGNAL; DECAY TIME IS APPROX. 600 ns
Measurements using an RF network analyser show that even a simple straight section of GIS exhibits a very complex transmission function (frequency-response curves). An example is shown below in Figure 11.
FIGURE 11: FREQUENCY RESPONSE OF A SIMPLE STRAIGHT SECTION OF GIS (0 – 2000 MHz)
In summary, as stated in CIGRÉ TB 444 [5]: “The measured amplitude and waveform of UHF PD signals is dependent on the location of the originating PD source, the distance to and position of the sensor, and the extremely complex RF propagation path between them. These frequency dependent effects are related to the dimensions and geometric shapes of the HV equipment (GIS).”
Page 21
Guidelines for PD Detection
Therefore it is impossible to directly and quantitatively link received signal strength of UHF PD signals to the actual charge level of the PD source. UHF PD SENSORS Although at first glance the GIS enclosure would seem to form an ideal Faraday cage, GIS enclosures do not completely screen out all external EMI. Obvious openings occur at overhead line (OHL) bushings and at cable end terminations, but the enclosure is also interrupted for current transformers, drives, flexible couplings and observation windows. Some leakage even occurs at compartment boundaries (owing to the skin effect at UHF frequencies).
Despite this, the GIS enclosure does provide a high level of RF shielding, and for this reason RF PD sensors are installed internally to achieve high sensitivity to the weak PD signals and minimize exposure to external EMI.
FIGURE 12: EXAMPLES OF UHF PD SENSORS
Typically RF PD sensors consist of a metallic electrode installed inside the GIS enclosure, connected to the outside world via a gas-tight RF coaxial jack. They are designed to pick up as much of the PD signal as possible while at the same time not compromising the HV design (indeed, this is a fundamental design conflict for GIS PD sensors). GIS RF PD sensors are sometimes referred to as antenna, but this technically incorrect usage and should be discouraged, since the term ‘antenna’ generally refers to a radiating structure operating in the far field, i.e. >10 λ. Again, the internal dimensions of GIS – indeed, the dimensions of UHF PD sensors themselves – are of the same order as the wavelengths at which we are measuring PD in the UHF region. Therefore GIS PD sensors are operating in the extreme near-field and are thus not ‘antennas’ in the strict sense of the word. The result is that the RF PD sensor is an integral design of the GIS; in general sensors are not interchangeable between manufacturers or even between different types from the same manufacturer. For this reason it is not very useful to try to make comparisons based on equivalent antenna height or other sensitivity criteria normally applied to RF antenna; removing an RF PD sensor from its specified OEM location totally changes the way it interacts with the PD signal for the reasons cited above. Should it be desired to verify the sensitivity of a UHF PD sensor, one suggestion would be to simply plot the signal from a moving particle of 5 pC magnitude (e.g. during the CIGRÉ Sensitivity Verification Method Laboratory Part 1) on a spectrum analyser together with a second sweep showing the background noise spectrum without high voltage. See Figure 13 below for an example.
Page 22
Guidelines for PD Detection
FIGURE 13: PD SENSOR SENSITIVITY SHOWN USING REAL PD SIGNAL (MOVING PARTICLE), HIGH-QUALITY PRPD PLOT (R)
In GIS which are not equipped with integrated internal UHF PD sensors, it is often possible to employ specially designed external sensors [26]. External sensors are designed to attach to the outside of the enclosure where there are openings which allow the RF PD signals to be picked up, such as observation windows or the openings in insulators through which the liquid dielectric is cast. Figure 14 shows an example of such an external RF sensor used in this way. Such external sensors can also be used to look for PD in e.g. GIS-cable terminations, which often have a section of exposed bushing through which some signal energy will leak out and be picked up.
FIGURE 14: EXTERNAL UHF COUPLERS FOR USE AT GIS SPACER CASTING INLET
The biggest problem with external RF PD sensors is they are less sensitive to PD signals (inside the GIS) but more susceptible to external interference (EMI). To mitigate the latter problem, external sensors may be installed using some sort of EMI gasket or shielding. In addition, the openings such as observation windows or casting inlet passages may be quite small (< 50 mm diameter); smaller size apertures reduce the amount of signal energy available for the sensors to pick up, and also usually raises their lower cut-off frequency. PD SOURCE LOCALIZATION USING THE TIME-OF-FLIGHT (TOF) TECHNIQUE Once a signal has been verified or is suspected to be PD, it is usually attempted to locate it, at least to within one GIS compartment. Since the 1980s the so-called ‘time-of-flight’ (TOF) technique has been used [6], [7], [14], [27]. This consists of connecting the output of two PD sensors to the inputs of a fast digital sampling oscilloscope (DSO), using high quality RF pre-amplifiers and RF coaxial cables of the same length and type, see Figure 9. (While it is possible to make TOF measurements using cables of different lengths, this should be discouraged because it can make the location calculation very confusing. If it cannot be avoided, one way to deal with such a situation is to measure the difference in time-of-arrival for the two cables using a fast pulse generator (e.g. as used for the CIGRÉ Sensitivity Verification) connected to their respective ends.).
Page 23
Guidelines for PD Detection
The DSO is used to determine the difference in the time-of-arrival of the PD signals from two sensors from which the location of the PD source is calculated following principle shown in Figure 15. The PD source will be located away from the midpoint between the two sensors, toward the sensor from which the signal arrives first. Note that the velocity of propagation in GIS is not the same as in a vacuum! Also, if one of the measurement paths contains more solid spacers than the other, this needs to be accounted for. The maximum possible spatial resolution is directly dependent on the time-domain resolution of the DSO and therefore on both its bandwidth and its sampling rate. For good location accuracy, the DSO input bandwidth should be at least 1 GHz and its minimum sampling rate 5 GS/s.
FIGURE 15: PRINCIPLE OF TIME-OF-FLIGHT (TOF) LOCALIZATION
In Figure 15, t is the difference in time-of-arrival in ns, d & D are in meters.
Being able to precisely locate a PD defect to within one compartment or even on which side of a flange can make a big difference in the time it takes to open a particular section of GIS to remove the PD source. However, obtaining high precision can be very challenging for various reasons. Weak signals and/or strong background EMI sources can mean poor SNR which may make determining the exact start time of the incoming signals difficult. Filters can be tried to reject the external EMI, but filters always exhibit group (time) delay effects which can introduce error in the calculation. If filters are used, they should be employed in pairs (one for each connection) and they should be closely matched in their group delay characteristics. If only one filter is available, the difference in delay time between the two should be checked and precisely quantified using the method for unequal cable lengths described above. Another potential source of error in TOF measurements can arise when determining the midpoint between the two sensors as exactly as possible. Usually this is done e.g. with a tape measure, but it takes experience to know which ‘line of sight’ distance to measure along the GIS measurement path. VERIFYING THE SENSITIVITY OF PD MEASUREMENTS USING UHF TECHNIQUES As has been well-known since the earliest work in UHF PD methods, and as previously stated, it is not possible to correlate the apparent charge of a PD source with the received UHF-signal strength. In other words, it is not possible to ‘calibrate’ PD measurements done using UHF techniques [21], [20], [28]. However, in 1999, CIGRÉ Task Force 15/33.03 published “Sensitivity Verification for Partial Discharge Detection System for GIS with the UHF Method and the Acoustic Method” [21]. This document has found wide acceptance throughout the GIS manufacturer and utility community and is increasingly applied as a de facto standard, often being cited in technical specifications for PD testing and monitoring of GIS. Only a very brief summary of the method is mtorsaade here for reasons of completeness; the reader is referred directly to [21] for more information.
Page 24
Guidelines for PD Detection
The method consists of two parts, a laboratory part and an on-site part, see Figure 16 below. In the first part, a section of GIS is embedded in a conventional IEC 60270 test circuit [13] which includes 2 compartments each of which is equipped with one UHF PD sensor. A real PD defect (typically a moving particle) is placed in the first compartment and adjusted such that it registers 5 pC charge magnitude. The UHF signal measured at the sensor in the second compartment and recorded. The high voltage is turned off, and a pulsed signal injected into the sensor in the first compartment which contained the real PD defect. This signal is adjusted until it matches the signal from the actual PD defect as closely as possible. In the second part, the signal from the same pulse generator using the same settings is injected into the sensors on site. If the signal can be observed at the next adjacent sensor(s), the verification check is considered passed for the section of GIS between those sensors.
FIGURE 16: CIGRÉ SENSITIVITY VERIFICATION: LAB PART 1 (L), ON-SITE PART 2 (R)
2.5
Case Studies
2.5.1 Case Study 1: External signals trigger alarms on UHF PD monitoring system An on-line UHF PD monitoring system installed on a 400 kV GIS was producing alarms based on signals picked up from UHF PD sensors installed in several sets of feeders connected to step-up transformers at a large power plant. Analysis of the PRPD patterns displayed by the PD monitoring system from all three phases show an obvious 120° phase offset and similar form; this strongly suggests the PD signals originate outside of the GIS – e.g. from within the step-up transformers – and are being coupled into the GIS PD monitoring system. Test equipment was brought out to several of the feeders in question and the results shown below. PRPD patterns from the PD monitoring system are shown in Figure 17, which also includes a PRPD pattern taken directly at PD the sensor in question for comparison.
Page 25
Guidelines for PD Detection
FIGURE 17: PD MONITORING SYSTEM DISPLAY SHOWING 120° OFFSET IN SIGNALS, Y-PHASE PATTERN FOR COMPARISON
TOF measurements were carried out at one of the feeders as shown in the detail on the GIS single-line diagram in Figure 18 (left). On the right, a screen-shot from the first TOF measurement shows the signal from the sensor closest to the transformer (blue trace) arriving 24.5 ns before the signal recorded on the next available PD sensor (red trace).
FIGURE 18: SINGLE-LINE DIAGRAM SHOWING PD SENSOR LOCATIONS (L), TOF MEASUREMENT (R)
To further confirm the signals’ origin in the step-up transformers, a second TOF measurement was made based on the principle shown in Figure 19. The measurement set-up was left exactly as it was as in Figure 18, but a fast GIS pulse generator was used to inject pulses upstream and outside the original TOF path. The result is shown in Figure 20. Note that the difference in time-of-arrival of the two signals remains the same, but the order of their arrival reverses, as would be expected, thus further confirming the PD signal source in within the transformer. Note also that the actual PD signals exhibit significantly lower frequency content than the signal
Page 26
Guidelines for PD Detection
from the GIS pulser; this is because the oil-gas bushing on the transformer acts as a low-pass filter which strongly filters out the higher-frequency components of the PD signals originating from inside the transformer.
FIGURE 19: METHOD FOR VERIFYING TOF PATH DELAY TIME AND DIRECTION USING AN EXTERNALLY INJECTED PULSE
FIGURE 20: TOF PATH DELAY VERIFIED WITH INJECTED PULSE; ALSO NOTE THE ORDER OF ARRIVAL REVERSES
Compare to Fig. 19 and note that CH2 now arrives 24.5 ns before CH3.
2.5.2 Case Study 2: TOF location of PD recorded by UHF PD monitoring system A UHF PD monitoring system installed in a 400 kV GIS was indicating sporadic, low-amplitude signals whose patterns, shown below in Figure 21, suggested a possible defect in a solid dielectric insulator. A PRPD pattern obtained using the UHF method (VIDEO OUT at zero-span) in Figure 9 shows good comparison.
Page 27
Guidelines for PD Detection
FIGURE 21: PD MONITORING DISPLAY SHOWING PATTERS SUGGESTING A DEFECT IN SOLID INSULATION
FIGURE 22: PRPD TAKEN AT THE SAME SENSOR USING UHF METHOD (ZERO-SPAN)
A series of time-of-flight (TOF) measurements was made in the area of the GIS containing the sensor which was picking up the signals as shown in the drawing on the left-hand side of Figure 23. The area of interest was narrowed down to the two sensors shown in the photo on the right-hand side. A plot showing the time-of-arrival of the two signals is shown in Figure 24. Note that the separation is only 390 ps. The signal arriving first (red channel) is from the lower sensor shown in the photo in Figure 23. The insulator was replaced and the signal disappeared.
FIGURE 23: LOCATION OF SENSORS USED FOR TOF MEASUREMENT (L); PHOTO OF SENSORS –Q56Y & PD30Y (R)
The arrow indicates the insulator which was removed (the PD signal subsequently disappeared).
Page 28
Guidelines for PD Detection
FIGURE 24: PLOT OF TOF OF FIG. 21: RED TRACE IS THE SIGNAL FROM THE LOWER SENSOR (PD30Y)
The insulator was exchanged and the signal disappeared. Later X-ray analysis on the insulator revealed a small void near the centre conductor (dia. approx. 700 µm). 2.5.3 Case Study 3: TOF location of PD signals in a cable bushing Once again a UHF PD monitoring system was indicating a strong source of PD in a feeder on a 400 kV GIS. The PRPD pattern, shown in Figure 25, suggested surface discharges which is often associated with bushings.
FIGURE 25: UHF PD MONITORING SYSTEM DISPLAY SHOWING PRPD SUGGESTING SURFACE DISCHARGE ACTIVITY
Measurements employing UHF techniques were quickly made at several sensors along the feeder which indicated the source of the signals to be at the exit, outside of the building. A TOF measurement was set up on the GIS section shown in the single-line diagram in the left-hand side of Figure 26. The right-hand photo shows an external UHF sensor and pre-amplifier attached to the outside of the cable bushing. A plot showing the difference in time-of-arrival of the signals from the two sensors is shown in Figure 27. This time delay corresponded closely with the actual distance between sensors 1 & 2 and was subsequently verified using the technique diagrammed in Figure 19.
Page 29
Guidelines for PD Detection
FIGURE 26: TOF SET-UP TO LOCATE CABLE BUSHING PD, PHOTO OF EXTERNAL SENSOR #1 (R)
FIGURE 27: PLOT OF TIME-OF-ARRIVAL OF TOF DIAGRAMED IN FIGURE 26 ABOVE: APPROX. 38 ns
2.5.4 Case Study 4: External EMI recorded by UHF PD monitoring system This case study shows measures taken for a problem often encountered with GIS PD monitoring systems employing the UHF technique: outside signals entering via openings in the GIS enclosure (previously described) and being picked by the PD monitoring systems, sometimes even triggering alarms. The first is a PD monitoring system installed on a 400 kV GIS in the Middle East; typical signals for the sensor in question are shown below.
Page 30
Guidelines for PD Detection
FIGURE 28: UHF PD MONITORING SYSTEM DISPLAYING SIGNALS STRONGLY SUGGESTING EXTERNAL EMI
The two strongly defined, vertical lines in the left-hand pattern are typical of causal (i.e. modulated) signals.
Measurements were made with an RF spectrum analyser in the area of the PD sensor from which the PD monitoring system was recording the suspect signals. Plots of what was found are shown below in Figure 29.
FIGURE 29: MEASUREMENTS OF EXTERNAL EMI PICKED UP BY A UHF PD MONITORING SYSTEM
The plot on the left shows the spectrum of the signal which is typical for a radar pulse. One the right, the signal is down-converted to show the modulation on an oscilloscope in the time domain.
In another example, the CIGRÉ TF 15/33.03.05 Sensitivity Verification was being carried out during commissioning testing of an on-line UHF PD monitoring system. While injecting the external test pulse, a signal with very similar appearance found seen slowly ‘walking’ through the display as shown in Figure 30 (this is typical of signals which are not quite synchronized to the line frequency). This led to obvious confusion for that sensor. Investigation was under-taken, and the signals are shown below in Figure 31, the spectral appearance but especially the oscilloscope plot showing the obvious digital modulation demonstrated these were causal signals. Finally the source of the signal was traced to the remote control of the building crane – verified by switching it off. This is typical of the type of troubleshooting required to identify external interference signals which are getting coupled into the GIS and being picked by UHF PD monitoring systems.
Page 31
Guidelines for PD Detection
FIGURE 30: LOW-LEVEL PULSED EMI INTERFERING WITH CIGRÉ SENSITIVITY VERIFICATION CHECK
Two photos were made of the display using a hand-held camera; along with a PRPD pattern of the EMI (r).
FIGURE 31: CAUSAL EMI, SPECTRUM (LEFT), OSCILLOSCOPE PLOT SHOWING DIGITAL MODULATION (RIGHT)
2.6
Advantages and disadvantages of different methods
TABLE 3: COMPARISON OF UHF AND CONVENTIONAL (IEC 60270) METHODS (GIS)
Comparison of conventional (IEC 60270) vs. unconventional (IEC 62478) PD methods applied to GIS Conventional
UHF Method
Acoustic methods
Calibrated in pC
Charge calibration not possible; CIGRÉ sensitivity check
Charge calibration not possible
High sensitivity (<1 pC)
High sensitivity (< 5 pC)
High sensitivity (< 5 pC)
Coupling capacitor (CC<
Built-in & external RF sensors
Typ. piezoelectric acoustic energy
EMI difficult to suppress
EMI relatively unproblematic
No EMI background vibration
Not possible to locate PD
Allows implementing time-of-flight Excellent for locating many types of PD (manually or TOF) (TOF) method for locating PD
Type testing, factory final test
On-site commissioning, monitoring
Page 32
assessment,
Useful for both factory and on-site
Guidelines for PD Detection
2.7
Summary and important aspects
1. Once again it is emphasized that when making PD measurements using unconventional methods, it is not possible to quantitatively correlate or calibrate the received signal level in terms of PD charge(pC or nC). 2. The main emphasis when employing unconventional techniques for PD detection and localization is to obtain the best SNR possible through careful development of sensors and utilization of appropriate signal processing techniques. The performance and behaviour of each component in the entire signal-handling chain needs to be well-understood in terms of its effects on the PD signal. 3. When performing unconventional PD measurements, verification checks should be made whenever possible, especially during on-site measurements, to assure that the measurement system components are all performing in the way expected. This follows the concept of ‘best practise’ when making conventional (IEC 60270) PD measurements in the lab or in the factory: when changing test objects or adjusting the set-up, we take a quick calibration check. Each component in the measurement chain should be checked periodically to prevent overlooking e.g. a bad connector or a faulty pre-amp, again because there is no general calibration procedure as in IEC 60270 which might catch such problems in the signal path(s). 4. Again it is good to follow general best practise procedures, i.e. extensively documenting test set-ups, taking detailed notes, photographs, and so on. This is especially important when employing unconventional methods because the test set-ups are not industry standardized, but typically assembled according to the particular equipment under test and the needs of the user. PD measurement and monitoring of GIS based on the UHF method is a topic of great interest both in the worldwide GIS community and in CIGRÉ SC D1; the activity presented here is a response to that. The first step of a risk assessment procedure - to perform on-site PD measurements and detect critical defects in GIS - was proposed in CIGRE TB 525. In addition, implementation of the UHF method has been further improved in close cooperation with the activity of WG D1.25 with the publication of CIGRE TB 654, “UHF partial discharge detection systems for GIS: Application guide for sensitivity verification”, the revised and updated version of CIGRÉ TF 15/33.03.05 from 1999. During the past 20 years or so, manufacturers and users of GIS have been gaining more experience with the application of partial discharge monitoring systems using the UHF method for PD monitoring in service. However, the technical requirements of PDM systems are not standardized, and the related recommendations, derived from the different application cases, vary significantly. There is a strong need for versatile PD monitoring systems to be able to provide usable data, in line with the relevant international documents, which is required to assure that actual PD defects are detected and their evolution is monitored so that asset managers can derive prompt actions. However, the relationship between the information (signal data) provided by the monitoring systems and the actual risk presented by a given PD defect is still not exactly defined. This challenge is faced by both the manufacturers and users of GIS and the PD monitoring system manufacturers, and much work still remains to improve understanding in this area.
Page 33
Guidelines for PD Detection
3 POWER CABLE SYSTEMS 3.1
Typical PD sources in power cable systems
Partial discharges in the power cable insulation system can be caused by cavities, contamination, or defects at interfaces. The electrical stresses in the cavities are dependent on the shape and location of the cavity. The breakdown strength of a cavity depends on its dimensions along with the composition and pressure of the gas inside. Gas-filled cavities may also be formed by local breakdown sites due to field concentrations around conducting inclusions. Furthermore, treeing can be initiated from an inclusion in solid materials or discharges in oil can occur from inclusions. Long lengths of cable behave like a transmission line and cannot be treated as a lumped capacitance for PD analysis. If a partial discharge occurs in a cable system, the resulting fast current pulse will result in a travelling wave which propagates away from the original location. At the measuring side of the cable (impedance: Zdetector >> Zcable), a reflection of the PD wave will occur. The PD signal wave will travel to the remote end of the cable system (open during measurements) where a full reflection occurs again. This process repeats until the PD wave is attenuated, partially due to reflections at impedance changes (e.g. joints), but also due to attenuation and dispersion during propagation through the cable insulation system [29]. In newly installed cable systems, partial discharges can be introduced due to failures in the production of the cable components or of the assembly of the cable components on-site [30]. For cables systems in service, different stresses such as electrical, mechanical, thermal, or chemical, singly or in combination, may cause PD-sources. The type of PD-source in the cable insulating system can be determined by analysis of the PRPD pattern which is recorded during the PD measurement. The PRPD patterns are reflecting the physical phenomena (stochastic behavior) of the specific PD-sources. A starting electron is necessary for PD activity to begin within a local region as defined above. The availability of starting electrons is strongly dependent on the PD source itself, the insulation material (solid, fluid or gas material), and on the position of the PD source with respect to the electrodes where the electrical field strength is present. The comparison of PD detection and PD results evaluation at different voltage shapes e.g. DAC, VLF and 0.1 Hz is extensively discussed in [29]. Investigating the same PD-source using different measurement principles, e.g. conventional electrical (according to IEC 60270) along with unconventional e.g. HF measurement methods, theoretically a similar PRPD pattern may be obtained, given the sensitivity of the measurement system is sufficient, the PD pulses from the defect have sufficient amplitude and the ambient noise influences are not too strong. As the occurrence of partial discharges is a stochastic process, and the detection can be influenced by several factors, typical PD patterns appear in many variations. Therefore experience and interpolation / abstraction capability is required for correct interpretation and screening of the PD type based on the measured PD pattern.
3.2
PD pulse propagation
Because the length of cable systems can range up to several kilometres and because of the influences of disturbances and noise during onsite measurements, pulse propagation is an important parameter for partial discharge measurements. Because of the inherent dynamic range of the measurement devices, e.g. digitizer or oscilloscopes, the signal amplitude of the measured pulse must be sufficiently high enough, otherwise the PD
Page 34
Guidelines for PD Detection
signal disappears in the background noise level. Overall measurement sensitivity is fundamentally important in partial discharge measurements. If the measurable pulses are too small or the background noise level is at too high level, reliable information about the condition of the insulation cannot be obtained. Signal processing tools for noise suppression are used to improve the PD sensitivity in high disturbance environments such as often encountered during line PD measurements. Large test objects like power cable systems have significant influence on PD pulses, due to attenuation and pulse deformation caused by dispersion. In particular the sensitivity of the measurement will also be influenced by the geometrical dimensions e.g. lengths, conductor diameter and insulation thickness of the particular cable under test. Not only the magnitude of the pulse will decrease by travelling along the distance of the cable, also the shape may be modified. Therefore it is advantageous to have knowledge about the influence of these factors on PD pulses in such cables. Different parameters influence the propagation of a PD pulse in the time domain. These parameters differ for each cable type. The input parameters are attenuation and wave propagation velocity; both parameters are frequency dependent as illustrated in Figure 32.
FIGURE 32: FREQUENCY-DEPENDENT PHASE VELOCITY AND ATTENUATION PARAMETERS USED FOR CABLE MODEL CALCULATION [31]
The third parameter which is necessary for the calculation is the distance travelled by the pulse. This parameter differs for each unique PD source origin and is an important factor for the calculation. For each distance, the model needs to recalculate the pulse shape and pulse amplitude. Furthermore, the influence of frequency-dependent attenuation and signal dispersion are dependent on the distance travelled by the pulse for the specified cable length. The length of a power cable system is an important limitation of a reliable PD diagnosis. Due to attenuation and dispersion, the measureable signal decreases with the distance travelled and the length of the cable. The attenuation decreases the amplitude with increasing travelling distance. Also, dispersion leads to different results with longer travelling distance, as a result of impulse widening and increase in the rise time as compared to the integration time constant. Moreover, dependent on the actual noise levels and the finite dynamic range of the measurement equipment, the overall sensitivity is affected. Due to these factors, a maximum distance between a PD source of certain magnitude and a measurement position can be defined to guarantee an accepted sensitivity of PD measurements.
Page 35
Guidelines for PD Detection
3.3
Conventional electrical PD measurement according to IEC 60270
PD pulses generated within the cable system as well external occurring PD sources/disturbances can be typically measured by using an external coupling capacitor. For this purpose the coupling capacitor is connected to the cable termination through a coupling device which is in turn connected to a measurement instrument at one end of the cable. The typical set up is shown in Figure 33.
v
A
B
v
Lcable Ck CD
Components MI
Ck
Coupling capacitor
CD Coupling device MI
Measurement instrument
FIGURE 33: TEST CIRCUIT FOR MEASUREMENT WITH A COUPLING CAPACITOR AT THE CABLE TERMINATION
The test capacitance of the cable depends on the type of cable and its overall length. The application of the conventional PD measurement according to IEC 60270 has several advantages:
Calibrated readings of apparent charge value in [pC].
Reference PRPD patterns for different discharging defects and components are available (well-known method since the 1960s).
Standardized procedures exist for type-, factory- and on-site tests. According to IEC 60270, it is recommended to measure the apparent charge by using a defined wideband or narrowband measurement system. The PD measurement system is connected via a coaxial cable to the measuring impedance CD. The charge q, measured in units of pC, corresponds to the charge transferred during the voltage drop between the couple capacitor capacitance Ck and the test object capacitance. The voltage drop (and the corresponding charge transfer) can be triggered by an external event or an internal PD source. Because a cable is a distributed object with a certain length, the PD pulse needs to travel a certain distance to the coupling capacitor. The readings of the PD measurement are influenced by the position of the defect, see Figure 34 [32].
Page 36
Guidelines for PD Detection
FIGURE 34: MEASURED AND SIMULATED PD MAGNITUDE AS FUNCTION OF TRAVELLING DISTANCE, NORMALIZED TO 500 pC FOR THE FIRST VALUE
To know the attenuation of the PD signal along the cable, a calibration of the test circuit must be done. For this reason, a reference charge is injected from a calibrator connected between the high voltage terminal and the ground (cable sheath) of the cable termination. The PD measurement system is adjusted in accordance with the injected reference signal. The first PD measurement should be made at a low test voltage level (e.g. 10 % of the rated voltage) in order to determine the background noise level. The maximum background noise level should be less than 50 % of the maximum admissible apparent charge value specified for the cable under test. However as the noise level at the on-site test location cannot be predicted and is influenced by many factors, the on-site conditions must be taken as given. This can result in high noise level readings due to electro-magnetic interference (EMI). The PD level must be checked at each cable termination of the three phases (single-ended). The PD test is considered successful if no continuous PD activity is detected with values above the maximum admissible level specified at any cable termination and if there is no rising trend in the apparent charge amplitude during the withstand test. If a PD event occurs, a deeper investigation of the PD event shall be done. A PRPD analysis accumulates information during several cycles of the voltage, which is important for identification of the type of PD source. A typical example of a PRPD pattern is shown in Figure 35.
Page 37
Guidelines for PD Detection 30
Voltage (kV)
20 10 0 -10
0
0.5
1
1.5
2
2.5
3 Time (ms)
3.5
4
4.5
5
5.5
0
0.5
1
1.5
2
2.5
3 Time (ms)
3.5
4
4.5
5
5.5
-20 -30
600 500 PD (pC)
400 300 200 100 0
FIGURE 35: TYPICAL EXAMPLE OF A PD PATTERN MEASURED DURING AN ON-SITE CABLE TEST
The on-site application of conventional measurements according to IEC 60270 for PD measurements on power cable systems can have several disadvantages:
In case of high levels of EMI on-site, it may be difficult or impossible to obtain satisfactory noise level readings, and consequently, no reliable apparent charge value of an internal defect in the cable under test can be obtained.
The length of the cable and the type of cable have influence on the sensitivity of the measurements.
The relatively low measurement frequency bandwidth makes the method unsuitable for PD source localization.
The PD magnitude is strongly influenced by defect position, particularly for long cable lengths.
Measurement sensitivity is reduced in the case that the capacitance of the coupling capacitor is low compared to the cable capacitance.
The on-site HV power source used can also sometimes be a source of EMI noise.
PD measurements on long cable lengths require measurements at both cable ends.
3.4
Unconventional HF PD-measurement with coupling capacitor
The conventional PD measurement with coupling capacitor can also be performed in an unconventional way as a high frequency (HF) measurement. The PD measurement bandwidth is increased up to e.g. 50 MHz, which allows the higher frequency content of the travelling PD pulses to be utilized to localize the origin of the PD. A commonly used method to do this is time domain reflectometry (TDR). Using this technique, the reflection of the PD pulse from the far end of the cable has to be detectable by the measurement device. 3.4.1 Single-ended measurement and PD origin localization The classical single-ended measurement technique uses one PD detection circuit. A schematic setup is shown in Figure 36, which contains a PD detector, a coupling capacitor and a cable. On the left side of the cable, the inner conductor of the cable is connected to the coupling capacitor. The far end of the cable has an open end. Figure 36 shows a typical setup for energizing and measurement of PD in a power cable.
Page 38
Guidelines for PD Detection
FIGURE 36: MEASUREMENT SETUP FOR SINGLE SIDED PD MEASUREMENT
At the defect location, every PD generates a certain amount of electric charge. This process results in two impulses in the cable travelling away from the defect origin in each direction (Figure 37), propagating through the cable. At the moment tA the detector recognizes the first pulse A which has travelled directly from the PD origin to the near end of the cable. A second pulse B can be measured with a time difference ∆t. This pulse travelled from the PD location to the far end, was reflected, and then travelled back in the other direction to the near end, to be measured a certain time ∆t later. This time difference ∆t is crucial for localizing the PD origin. To calculate the distance XPD between the measurement device and the PD origin, equation (3.1) is used, where Lcable is the overall length of the cable and v the pulse propagation speed which is dependent on the cable characteristics. X PD Lcable t v 2
(3.1)
Especially in noisy environments or with long cable lengths, this technique has limitations in locating the PD origin. In particular, the pulse reflected from the far end has a longer distance to travel before reaching the measurement device; this added attenuation and dispersion consequently decreases the pulse’s amplitude. Depending on the specific cable characteristics, there is a limitation for the localization of the PD origin with increasing cable length. The sensitivity of the PD localization also decreases with increasing background EMI. Disturbances can occur even if the cable system is not energized. This sensitivity also depends on the effectiveness of the noise suppression tools implemented.
FIGURE 37: PD SOURCE LOCATION BY TDR ANALYSIS OF THE TWO TRAVELLING WAVES A AND B WITH SPEED V
Page 39
Guidelines for PD Detection
This TDR technique is based on the measurements of the first incident PD pulse and its reflection from the opposite cable end. It can be possible that the reflected pulse is not detectable anymore at the left end where the PD detector is located. To overcome this problem, the double-ended measurement can be performed. This technique increases the detection distance in which the PD origin localization is possible. As an example, partial discharges occur at the near end of the cable where the PD detector is located. The first measurable pulse appears with a fast rise time and high amplitude compared to its reflection. The second detectable pulse is time delayed by two times the total cable travelling time, referenced to the specific wave propagation velocity. In Figure 38, the PD is located at the near end. After the detection of the first pulse, the reflection occurs after circa 3.5 µs. The amplitude decreased due to attenuation while travelling through the cable. Also the rise time is slower compared to the first incident pulse. To calculate the location of the PD source, the speed of the PD pulse in the cable together with the time difference between the first and the reflected PD pulses is used.
FIGURE 38: NEAR-END PD FAULT LOCATION WITH REFLECTION FROM FAR END ON A 500 m CABLE
3.4.2 Double-ended measurement and PD origin localization Detection and localization of PD in cable systems with longer lengths can be improved by performing PD measurements at both ends of the cable circuit, as shown in Figure 39. This will, for the worst case situation (PD at the near end), reduce the required travelling distance for the reflected PD pulses by a factor of 2. In the single-ended measurement setup, a near-end partial discharge must travel through the whole cable length to the far end and the whole cable length back to the near end. The overall travelling distance is therefore almost two times the cable length. In double-ended measurements, the near-end PD only has to travel to the far end to be detected there, thus the PD signal traverses the cable length only once. This setup requires a coupling capacitor with PD detector A on the left side and a coupling capacitor with PD detector B at the right side.
Page 40
Guidelines for PD Detection
FIGURE 39: SETUP FOR DOUBLE-ENDED PD MEASUREMENT WITH LOCALIZATION CAPABILITY
PD detectors are installed on both ends of the cable system as shown in Figure 39. The detectors have to be synchronized to correlate the measurement data of both ends. Figure 40 shows the principle setup and the distances used for the calculation of PD origin. The overall length Lcable of the cable system is divided into two parts, the distance from the left end to the PD origin XA, and the distance XB from PD origin to the right end of the cable. TA and TB are the times at which the direct incident PD signals are measured. Both PD detectors must be synchronized for PD origin localization.
FIGURE 40: SETUP FOR THE DOUBLE-ENDED PD MEASUREMENT WITH FAULT LOCALIZATION FUNCTIONALITY
Both measurement units record data during high voltage application; therefore the measurement settings and the measurement data are communicated between the two measurement units. Furthermore, the units are time synchronized to obtain phase-resolved PD patterns at both ends as well as synchronized localization analysis.
X PD Lcable 2 t v 2
To calculate the PD origin, equation (3.2) can be used. The distance XPD, is the distance from the left detection unit to the PD origin and ∆t is the time difference of tA and tB. In this particular configuration, the PD pulses are directly measured and there is no need to register the reflections as is the case with the single-ended TDR evaluation. Since the units are synchronized with each other, the difference in the arrival times of the pulses at both ends together with the pulse velocity obtained from the calibration provides the location of the discharging defect. The main advantage of the double-ended measurement is its increased sensitivity compared to the singleended measurement. The PD pulses have to propagate only to the ends of the cable, no reflection from the far end is necessary. As a result, there is no additional attenuation due to imperfect reflection or attenuation from travelling back through the whole length of the cable. Page 41
Guidelines for PD Detection
The attenuation results in a decrease of the amplitude of measureable partial discharge pulses. This attenuation can be described with the exponential function shown in Equation (3.3), with two constants c0 and c1 in the exponent and with PD0 as the initial PD amplitude, and x as the variable representing the length travelled.
1 1 a PD PD0 exp x c 0 1 c1
As can be seen from Figure 41 and Figure 42, the double-ended measurement results in a higher sensitivity of the PD detection than the single-ended measurement. As an example, a 10 km long cable is simulated with single and double-ended measurements. For the single-ended measurement, normalized PD amplitude of 100 percent for the near end pulse and a reflected pulse of about 4 percent could be detected. With a doubleended measurement setup, the remaining PD amplitude on the right end is about 19 percent. In the case of a fault location between both ends, an example is taken of a PD origin at 2 km seen from the left end (Figure 42). Due to the attenuation, the decreasing PD amplitude has about 70 percent of the original amplitude. In a single-ended measurement setup the reflected pulse over the far end decreases to e.g. 5 percent. A double ended measurement increases the remaining PD amplitude to about e.g. 26 percent.
FIGURE 41: NORMALIZED PD ATTENUATION FOR DOUBLE- AND SINGLE-ENDED MEASUREMENT (10 km).
The grey curve indicates the pulse reflection of the single sided measurement.
Page 42
Guidelines for PD Detection
FIGURE 42: NORMALIZED PD AMPLITUDE ATTENUATION WITH A FAULT LOCATION AT 2 km FROM THE LEFT END.
The reflection from the right end is shown in grey.
The double-ended measurement offers significant advantages in detection sensitivity and localization possibilities of PD pulses over long cable lengths. Due to the use of two PD measurement units, the hardware is more complex than compared to the single-ended measurement. In particular the synchronization and data transmission over long distances is a challenge for further investigations. As an example, a synchronized and combined double-ended measurement is shown. The measured data from each PD detector from the left and right end are plotted in one graph. Measured data from detector A is in black and the data from detector B in grey. The PD pulses are generated 220 m from detector A and 424 m from detector B. The first pulses of both measurements are the direct travelled pulses. After the double cable travelling time the reflections could also be observed.
FIGURE 43: DOUBLE-END PD MEASUREMENT DATA WITH PD ORIGIN AT 220 m FROM DETECTOR A AND 424 m FROM DETECTOR B
It can be seen that the double-ended PD measurement shows a definite advantage in higher sensitivity and better localization possibility. The higher sensitivity enables testing longer cables with the same sensitivity compared to single-ended measurement. Furthermore the difference between near-end and far-end PD could be easily discerned.
Page 43
Guidelines for PD Detection
3.5
Application of wideband HFCT for UHF PD detection at joints and terminations
Measuring PD signals in the VHF/UHF range presents some fundamental differences to traditional measurements below 20 MHz. In the range below app. 20 MHz, a high-frequency current transformer (HFCT) mounted on the ground strap of a MV or HV cable termination will show the same signal level no matter where it is positioned. However, at higher frequencies, resonance and reflection effects must be considered. 3.5.1 Application of HFCT at ground straps of terminations and joints If much higher frequencies are used, the physical dimensions of cable accessories with respect to the wavelengths being measured play a more important role. Terminations are quite long and frequently grounded with long earthing straps. The construction of a joint is a significant deviation from the geometry of the cable itself and will thus cause significant reflections at higher frequencies. When the dimensions of non-coaxial sections come into the range of a quarter wave length at higher frequencies (VHF/UHF1), the current distribution is varying quite significantly with the position along the object. A joint or termination will show such at frequencies according to Table 4. TABLE 4: TYPICAL DIMENSIONS OF JOINTS
The consequence of resonances in the joint section is that the voltage and current distribution of the signals are no longer homogeneous. Figure 44 and Figure 45 show a pickup for the H-field which was positioned at different points along the ‘earth strap’ of the simulated joint.
FIGURE 44: H-FIELD AT THE MIDDLE OF THE ‘JOINT’ SECTION 1
VHF = 30 to 300 MHz, UHF = 300 MHz to 3 GHz
Page 44
Guidelines for PD Detection
FIGURE 45: H-FIELD AT THE END OF THE ‘JOINT’ SECTION
There are significant variations in the measured signal amplitude depending on the position of the sensor, the mechanical dimensions of the accessory, and the frequency observed. The frequency range in Figures 44 and 45 ranges from 150 MHz to 850 MHz. There is no ideal or even compromise position where a reliable measurement of the propagating PD signal can be taken. If E-field sensors are used, the situation is more or less the same, just that the positions of minima and maxima are reversed. This holds true for wideband HFCTs and magnetic & electric field sensors. A sensor placed at the wrong position will pick up no signal or a signal below the detection threshold of the test instrument, although a PD signal is propagating along the cable. The level indicated can be off by up to 30 dB, making the test irrelevant. Wideband measurements are not that dramatically affected, as they integrate over a wider spectrum and that can equalize peaks and nulls to a certain degree. However, they are more prone to receiving man-made signals, e.g. such as GSM (mobile telephone), giving misleading or incorrect levels. Narrow band test with spectrum analysers or UHF detectors, especially in spot test, can show significant level errors. The test frequency is normally chosen at in a region of the spectrum where no man-made interference is present. However, there is no guarantee that a representative level of the PD signal can be measured at the spot chosen for the sensor. 3.5.2 HFCT application at cross bonded joints Cable systems with higher power and / or higher voltage are frequently built with cross bonding joints. As it’s difficult to access HV cables in operation and have suitable access points close to the joints, a simplified model was built for lab use. It uses only two cables (phases) and is made of standard lab coax cables (RG 58). Two such RG 58 cables 20 m in length represent each phase of the HV power cable, and two additional cables of 5 m length simulate the connection from the joint to the cross-bonding box. The cross bonding is done with two 4 m cables, on one of which a wideband HFCT is installed.
Page 45
Guidelines for PD Detection
FIGURE 46: MODEL FOR APPLICATION OF HFCT TO A CROSS-BONDED JOINT
As the ends of the HV cable are not normally accessible during operation, the only position to measure PD is at the cross bonding box. In the model setup shown in Figure 46, the current in the cross bonding link was measured with an HFCT. The HFCT used has a coupling attenuation of 21 dB (Rt = 4.5 ) and a bandwidth of about 200 MHz.
FIGURE 47: CURRENT LEVEL IN A CROSS BONDING LINK
The signal level in Figure 47 shows that it is basically possible to use the cross bonding links for VHF / UHF measurements. The black line, labelled ‘aHFCT’ represents the basic attenuation vs. frequency of the HFCT used, so the additional attenuation arising from the cross bonding is the difference between this black line and the green curve. In general, the signal level is reduced by 10 dB to 20 dB. If the frequency choice is not correctly made, e.g. 156 MHz in this example, the excess attenuation can reach up to 40 dB as well.
3.6
System calibration with sensitivity and performance check
3.6.1 Conventional electrical PD measurement according to IEC 60270 The goal of calibrating a PD measurement setup is to verify that the measuring system will be able to measure the specified PD magnitude correctly in accordance with IEC 60270 [13]. The calibration of a measuring system with the complete test circuit is performed to determine the scale factor k for the measurement of the apparent Page 46
Guidelines for PD Detection
charge. As the capacitance of the test object has influence on the circuit characteristics, the calibration has to be performed for each new test object. The calibration of the measuring system with the complete test circuit is performed by means of injecting shortduration current pulses of known charge magnitude from a calibrator connect to the terminals of the test object (cable termination). The calibration can be performed with different pulse magnitudes in the relevant range of the magnitudes expected. After the calibration the on-site noise conditions can be evaluated and the proper calibration range can be selected to assure a good accuracy for the specified PD magnitude. The applicable range of magnitude should be in the range of 50 % to 200 % of the detected PD magnitude [13]. Due to attenuation and distortion of travelling waves along cable lengths, the magnitude of apparent charge which is recorded at a terminal of the test object can differ in magnitude from that at the point where it originates. The magnitude recorded at the terminal of a cable under test can be influenced by resonance phenomena or by reflections at the terminals. Reflection phenomena in cables can be taken into account using special calibration techniques. The sensitivity verification and noise reading can be performed as shown in the example of Figure 48 where calibration pulses of 100 pC are injected in the test object. The response of the pulses should be according to the injected value with a deviation of ± 10 % for a single pulse [13].
FIGURE 48: EXAMPLE OF AN ON-SITE PD CALIBRATION, INJECTING 100 pC PULSES EVERY 10 ms
3.6.2 HF PD-measurement with coupling capacitor Based on knowing the pulse travelling time, the origin of the PD source can be determined. Not only the PD origin is an important information quantity for condition diagnosis, but also the PD magnitude is helpful information about the defect and its criticality. The determination of PD magnitude is calculated based on the signal which has travelled from the partial discharge origin. As a result of the propagation of the pulses through the cable system, the detectable charge and amplitude of the pulse decreases due to attenuation and dispersion of the PD pulse. Knowledge about pulse propagation is an important parameter for proper measurements. Due to limitations of the dynamic range of the measurement devices, e.g. digitizer or oscilloscopes, the signal amplitude of the measured pulse should be high enough; otherwise the signal disappears in the background noise level. Also the sensitivity is an important value in partial discharge measurements. If the measurable pulses are too small or the background noise level of the measurement is at a high level, reliable information about the condition of the insulation cannot be obtained. Appropriate signal processing tools for noise suppression can be used to improve the PD sensitivity, especially for on-line PD measurements. This discharge can be described by a simplified equivalent circuit for a single PD pulse as e.g. the discharge of the capacitance of a void in solid insulation. Under application of alternating voltage this capacitance will be recharged by the displacement current over the virtual series capacitance to the actual PD site. This injects a current and voltage pulse which travels away from the PD origin in both directions, towards the end of the cable. This travelling wave will be reflected and attenuated at all changes of the characteristic wave impedance along the cable system, caused by joints, change of cable type, or terminations. The pulse duration of one PD pulse varies from less than 100 ns to over 1 µs, depending on the travelling distance (attenuation, dispersion and low pass characteristic). The pulses propagate through the cable with a wave Page 47
Guidelines for PD Detection
velocity of about 140 m/µs to 180 m/µs, depending on the dielectric constant of the insulation along with the physical properties of the other materials used in the construction of the cable. Because of these properties, a superposition or repetition of the detectable PD pulse can occur. The pulse amplitude and shape can be deformed. This is of importance for the measurement of PD and the verification of PD performance, especially on cable systems with length of less than the half propagation delay between the PD origin and a characteristic impedance change (joints, terminations, change of cable type). The minimum required PD level for each travelling distance can be derived because the maximum length for PD detection of a certain magnitude is limited. To evaluate this issue, a sensitivity check is proposed to provide a possibility to ensure reliable diagnosis. The following example explains the points discussed above. The grey signal in Figure 49 is a measurement with a certain noise level. The first signal peak on the left is injected by a PD calibrator set to a level of 50 pC.
FIGURE 49: PULSE SIGNAL MEASUREMENT COMPARED WITH RESPONSES CALCULATED FOR DIFFERENT TRAVEL LENGTHS
This injected pulse travelled the direct path from injection point to the measurement device. After 7 µs the signal reflected from the far end is detectable. Calculated signal responses for five different travelling lengths are shown in five corresponding colours. The three left-most signal responses could be associated to the directly propagated signal, while the last two signal fragments are obscured within the actual noise level. Therefore it could be concluded; that a PD level of 12 pC could be detected in this cable system at a distance of 1200 m; lower level signals will disappear in the noise level of this example. The sensitivity of the minimum detectable PD amplitude can be evaluated by performing a basic calibration at the energizing site. A calibration pulse is injected in the cable. The calibration pulse travels through the cable and will be reflected at the open end of the cable. The pulse will be attenuated due to travelling through the cable. If the reflected pulse is detected at the coupling capacitor, the minimal detectable PD amplitude can be determined if a PD pulse is originating from the far end. Furthermore, the amplitude at which a PD source can be localized in the cable can be determined for the worst case situation, i.e. where the PD source is near the beginning of the cable. An example of such a calibration is shown in Figure 50. In this example a calibration pulse of 200 pC is injected into the cable. It can be seen that the reflection, which travelled two times the length of the cable, is still visible. This shows that a PD source with a magnitude of 200 pC can be localised based on the time difference between the arrival of the first pulse and its reflection.
Page 48
Guidelines for PD Detection
FIGURE 50: EXAMPLE OF A CALIBRATION WITH A PULSE OF 200 pC ON A 10,500 m LONG CABLE UNCONVENTIONAL PD MEASUREMENT WITH HFCT SENSORS
In Figure 50, the original injected pulse, the reflections that occur at impedance changes (cross bonding joints) and the reflection of the injected signal at the end of the cable are shown.
For unconventional coupling of PD signals, HFCTs are a frequently used sensor type. There are many different designs on the market, so a sort of guideline is presented here which should be useful for judging their performance. Although they look quite simple, there are several properties which need to be examined for their effectiveness. Coupling ratio is one of the first parameters to look at. It is normally specified as transfer impedance, giving the output voltage, generally at a 50 loadwhen a defined current is flowing through a conductor centered in the HFCT. Values of around 4 to 5 are common. Turns ratio or coupling attenuation can be used as well. Higher values mean better sensitivity, but there is compromise between the lower cut off frequency fL and the saturation current Isat. The signal-to-noise ratio (SNR) is not affected over a wide range of Rt, as the noise is equally reduced, so that lower signal levels can normally be compensated for by adding additional gain. When making comparison tests between phases, cable systems, or multichannel records, the tolerance of Rt needs to be looked at as well. If level differences are used for diagnostics, they should result from actual variations in the signal, not by variations in the coupling ratio of the HFCTs being used. The transfer impedance is frequently cited in data sheets. When a vector network analyser (VNA) is used for testing, the normal output format is the s-parameter S21 over frequency, as it can be seen in the screen shots following. It can be interpreted as the insertion loss in dB. All of these results can be related to each other by conversion factors, e.g. an HFCT with Rtransfer = 4 would have an insertion loss of 22 dB in a 50 system.
Page 49
Guidelines for PD Detection
The lower cut off frequency fL, is also an essential parameter for an HFCT. It’s a compromise between the ability to detect dispersed signals from long cables and tolerating noise pickup from mains. Mains harmonics are measured and regulated up the 40th harmonic (= 2 kHz @ 50 Hz). There is a tendency to extend the standards to higher harmonics because the use of switching mode power supplies, variable speed drives, and LED lighting, all of which have observably increased the harmonic content of the mains current in recent years. The low frequency cut-off does not present a steep roll-off, being only a single pole, high pass filter response produced by the HFCT’s inductance combined with the load resistance (normally 50 ). Additional filter circuitry can be integrated into the HFCT to get better noise suppression, but it might be more useful to have a single pole sensor characteristic, without ringing or overshoot, and to add further signal processing if more filtering is required. Most of the HFCTs on the market operate in a range from 50 kHz to 150 kHz. For special applications, when no dispersed signals need to be captured, much higher frequency ranges can be used with the benefit of getting lower noise levels.
FIGURE 51: LOW FREQUENCY RESPONSE OF AN HFCT
In Figure 51, the green trace indicates the transfer function and the phase response is shown in magenta. The lower cut-off frequency is 80 kHz (-3 dB, cursor 2) and at 10 kHz the attenuation is 21 dB. Signals at 2 kHz, the limit of the regulated range of harmonics, are down by -32 dB. The wideband response of a similar HFCT is shown in Figure 52.
Page 50
Guidelines for PD Detection
FIGURE 52: WIDEBAND FREQUENCY RESPONSE OF AN HFCT
The high (upper) frequency cut off fU The maximum frequency at which an HFCT needs to respond is dependent on the application. If just the magnitude of the partial discharge in pC is of interest, a high frequency cut-off of 10 MHz is adequate. It’s a single pole low pass filter as well and thus preserves the charge value to a reasonable extent. By integrating over the signal, the charge is measured correctly, even if the peak amplitude value is already affected. More important is a flat top and a smooth roll off, without overshoot. If time coincidence between channels is of interest or the amplitude of short PD pulses needs to be captured, higher values of fu are needed. Higher bandwidth is also required for discrimination of local / cable PD (switchgear) or if building of signal clusters is desired. For characterising rise-time of PD pulses, the rise-time of the HFCT itself will typically set the limit of what can analysed.
FIGURE 53: 12 ns SQUARE PULSE WITH 17 MHz HFCT (MAGENTA) AND WIDEBAND HFCT (BLUE)
In Figure 53, the influence of the HFCT bandwidth on a narrow impulse is shown. A pulse with a nearly square shape and approximately 16 ns pulse width is measured with an HFCT with a fu of approximately 17 MHz. Its output signal is shown in magenta. The blue trace is from a HFCT with a fu of more than 100 MHz. It’s clearly visible how the peak amplitude is reduced, actually by more than half; in addition, the rise and fall times are increased and the overall pulse shape is altered by the low bandwidth HFCT. The measured charge in pC is the same for both traces.
Page 51
Guidelines for PD Detection
Shielding is an important feature of an HFCT, which is often not well covered in data sheets and publications. However it has influence on the performance, especially at the higher frequency end. When an HFCT is not installed close to the RF grounding point, but there is a long length of grounding strap in between, the PD current will exhibit a voltage drop.
FIGURE 54: E-FIELD COUPLING OF THREE DIFFERENT HFCTS
Figure 54 shows the effect of pure E-field coupling on 3 differently constructed HFCTs. The measurement is taken with a VNA, as it has plenty of dynamic range and good noise suppression. The level of magnetic coupling is shown graphically by the black line for reference. The greater the distance between the black line and the E-field response, the better the HFCTs’ shielding. The worst of them (dark green) shows only about 16 dB difference between the current (inductively coupled) and the electrical field signal output at 10 MHz; that is only a factor of approximately 6. It could be argued that both signal components are caused by the PD signal, so it doesn’t matter which one is picked up. However, it defeats the purpose of an HFCT, which is to couple out the PD current signal. But it also causes a second problem, as the coupling factor becomes dependent of the polarity of installation, or in other words, the current direction. The inductively coupled signal reflects the polarity of the current, whereas the portion of the output due to electrical field coupling does not. This means the coupling ratio at higher frequency varies with the orientation of the sensor, if its shielding isn’t adequate. E-field coupling can also lead to more noise pick up and reduce the S/N for the current measurements. If an HFCT has adequate shielding, there is no influence of the orientation on the coupling properties. It can also be installed anywhere along the ground strap and will produce the same output signal. Saturation effects: When an HFCT is installed at the earth strap of a cable, saturation should be no big issue. The circulating currents there rarely exceed 20 A, however, a HFCT for this application should be able to tolerate > 50 A of mains current components, to be on the safe side. For installations around the HV cable itself, > 250 A current tolerance will be required, for bigger cross sections > 400 A are not uncommon. HV cables with high currents are generally too big to use standard HFCTs around the cable. Special large size HFCTs should address the problem of coping with currents up to 1 kA. The main effect of the mains current passing through the HFCT is, of course, that the ferrite material becomes saturated and the amplitude of the PD signals thereby reduced as well. The permeability of the ferrite material approaches zero in full saturation and therefore, the PD signal is no longer couple out.
Page 52
Guidelines for PD Detection
The saturation current Isat is defined as the value of the mains current which reduces the PD amplitude by 3 dB. Polarity (orientation) The HFCT should be clearly marked in terms of its polarity relationship, e.g. an arrow pointing towards the grounding point of the cable or earthing strap. Basically the orientation doesn’t seem to matter, but if more than one HFCT is used in a test and signals need to be compared in polarity as well, a clear common definition is needed. A positive current flowing through HFCT in direction of the arrow should produce a positive polarity voltage at the output. Performance test set up: For the verification of the performance of HFCTs it is necessary to have a suitable test environment which has no influence on the properties to be measured. At frequencies above 10 MHz wire loops are not adequate anymore and a test setup assuring reasonable RF performance must be created. Requirements for a HFCT set-up:
For an HFCT with e.g. 25 MHz bandwidth, the test jig should have a BW of 50 MHz, with headroom for further development, e.g. toward 100 MHz as a usable upper frequency. It should be well matched to 50 for using standard pulse- or signal generators, VNAs and test cables. Results should be reproducible. Low E-field component at the mounting position of the HFCT. Shielded or semi shielded to avoid radiation (EMC regulations) and noise pickup, as far as handling is not compromised. It should be as compact as possible.
The main requirement for a test jig is to generate a known current in a conductor which passes through the centre of the current transducer (HFCT), where the current is independent of the operating frequency. The construction in the form of a rectangular box with two open ends – a sort of 'tunnel' - ensures a low inductance return path. In the centre of the box is a brass post of 10 mm diameter, perpendicular to the top and bottom of the box; this carries the current through the centre of the HFCT under test. The impedance match at the BNC connector is achieved just by putting a 50 resistor in series with the current post. As the total loop inductivity of the setup through the post and back through the sides of the box is about 50 nH, this configuration is good enough for working up to 50 MHz. If the upper useable frequency should be 100 MHz, a matching circuit in the feeding section needs to be added, which is installed inside the case on top. With the matching circuit installed, the input reflection of the test jig is very low in magnitude and phase up to 100 MHz, as measured with a calibrated Vector Network Analyser (VNA). In Figure 55, the blue trace shows that the input reflection of the empty test jig is better than -40 dB up to 50 MHz and still around -30 dB*) at 100 MHz. The magenta coloured trace shows that the phase of the input reflection is about 10° at 100 MHz. These values imply that the entire signal supplied from the generator flows through the brass current post, as very little signal is reflected. Finally, the green trace shows the magnitude of the input impedance. The scales for impedance and phase are on the right hand side of the graph. The mechanical construction of the tunnel jig is shown in Figure 56. There is an additional polypropylene plate to make sure that the HFCT is not resting directly on the aluminium sheet metal, in order to reduce its capacity to ground.
Page 53
Guidelines for PD Detection
FIGURE 55: INPUT REFLECTION OF THE TUNNEL TEST JIG
FIGURE 56: TUNNEL TEST JIG SHOWN WITH HFCT INSTALLED FOR TESTING
When VNAs are used for transmission testing, the results from measuring S21 are scaled in dB. Frequently the performance of HFCT is given as a transfer resistance in ohms. The insertion attenuation from a 50 VNA test system can be easily converted to transfer resistance, see Table 5.
Page 54
Guidelines for PD Detection TABLE 5: DB TO RT CONVERSION Ao/dB
- 34.0
- 30.5
- 28.0
- 26.0
- 24.5
- 23.1
- 22
- 20.9
- 20.0
Rt/Ω
1
1.5
2
2.5
3
3.5
4
4.5
5
Ao/dB
- 19.2
- 18.4
- 17.7
- 17.1
- 16.5
- 15.9
- 15.4
- 14.9
- 14.0
Rt/Ω
5.5
6.0
6.5
7.0
7.5
8.0
8.5
9.0
10.0
FIGURE 57: FREQUENCY RESPONSE OF A WELL DESIGNED AND SHIELDED HFCT: GREEN = TRANSFER FUNCTION; BLUE = RETURN LOSS *) A
return loss of 30 dB corresponds to a VSWR of 1:1.07.
3.6.3 Sensitivity check of on-line PD measurements applied to MV grids In practice, one of the main difficulties of on-line PD measurements in MV grids is to know if the sensitivity of the unconventional PD detection system to be used is sensitive enough to detect PD activity between every two consecutive PD sensors installed along the MV grid [33]. The maximum value of signal attenuation is achieved when the PD source is located close to one of two consecutive PD sensors, because in this case the PD pulses have to travel the full distance between both sensors. In practice, a maximum signal attenuation of around 20 dB(1) can be considered enough to achieve an appropriate PD location when efficient noise filtering tools are applied. Considering the frequency range of travelling PD pulses through cable systems of the grid is not more than few MHz (e.g. 7 MHz), the signal attenuation caused by the grid should be less than 20 dB over the frequency range being measured. (1) Note: The 20 dB level of admissible signal attenuation means that at least 1% of one frequency component of any PD signal generated close to a PD sensor arrives at the next consecutive sensor to be detected.
An efficient method to determine signal attenuation of MV grid is injecting bursts of sinusoidal signals of the measuring frequency range, e.g. between 1 MHz and 7 MHz. A burst duration Tb and a specific time delay between two consecutive bursts Td, should be used, e.g. values of 0.25 ms and 2.5 ms respectively are appropriate. These sinusoidal bursts can be superimposed on the grid voltage through the capacitive coupling installed in the metal enclosed switchgear to detect the actual voltage presence (see Figure 58). Figure 58 shows a PD monitoring system of a MV underground 15 kV cable of 3457 m interconnecting nine MV/LV transformer substations to a MV substation (MVS) used as reflection centre of 15 kV (on the left), and to
Page 55
Guidelines for PD Detection
an HV substation (HVS) 220 kV/15 kV (on the right). A first PD measuring system is placed in the MVS (Reflection Center, n=0), but a sensitivity check had to be performed in order to determine in which MV/LV transformer substation a second measuring system should be placed. A signal attenuation analysis was carried out using sinusoidal signals from 1 MHz to 7 MHz travelling along the MV grid. The injected signals in the MV grid were acquired by HFCT sensors distributed along the grid, connected at the cable sheath earth connection of each MV/LV substation. The first HFCT sensor was installed in the same metal enclosed switchgear where the signal bursts were injected. Other HFCT sensors were installed at the metal enclosed switchgear of transformer station numbers 2 and 3, located at 748 m and 1078 m respectively. Signal attenuation for the frequency range between 1 MHz and 7 MHz was determined for discrete frequency values spaced 1 MHz from each other. The sinusoidal burst signal was measured using the same unconventional PD instrument and the same signal processing software to be used for PD monitoring systems in order to reproduce the actual conditions under which the PD signals will be measured after installation of the PD monitoring system. The PD instrument used applies a Wavelet filtering tool with statistical signal processing software to remove superimposed noise(Patent PCT # MX/A/2012/013690); as can be seen at the bottom of Figure 58 the filtered wave-shape (red signal) clearly discriminates the sinusoidal bursts from the raw data (green signal).
FIGURE 58: ON LINE TEST TO DETERMINE SIGNAL ATTENUATION BEHAVIOUR AT INCREASING FREQUENCY
At both transformer substations numbers 2 and 3, the same signal attenuation tendency was observed. The signal attenuation along the cable grid increases as the sinusoidal frequency increases (see dash lines in Figure 59), but resonance frequencies can appear (e.g. in the Figure 59 a resonance frequency around 4 MHz appears for the analysed MV grid). In Figure 59, the signal attenuation at transformer station number 2 varies from -11 dB to -16 dB, while the signal attenuation at transformer station number 3 varies from -17 dB to 28 dB.
Page 56
Guidelines for PD Detection
MHz
dB
FIGURE 59: SIGNAL ATTENUATION VERSUS SINUSOIDAL FREQUENCY AT TRANSFORMERS SUBSTATIONS 2 AND 3.
The signal was superimposed on the grid voltage of the MVS (n=0)
On the basis of the signal attenuation shown in Figure 59, and taking into account that the PD monitoring system used can detect PD pulses in any frequency range between 1 MHz to 20 MHz, no PD measuring system (MS) was required at the first two transformer stations (n= 1 and n=2). Therefore, the second PD measuring system could be placed in transformer station n=3, 1078 m from the reflection centre. 3.6.4
Sensitivity and performance check of unconventional PD system
GENERAL Partial discharge testing is a very sensitive method of detecting small imperfections in the insulation of extruded HV cables such as voids in the insulation shield layer. When applied during after installation tests, such test can detect any damage done during shipping or any problems created by the laying, joining, or terminating the cable.
The suitability of PD measurements essentially depends on the actual noise level on site and on the achievable sensitivity. Because PD impulses are subject to strong damping as they propagate along the cable, conventional PD detection at the cable end strictly leads to a strong decrease in PD sensitivity with increasing cable length. However, the cables already have been PD-tested during their routine test at the manufacturers, so cables should be free of PD faults when leaving the factory. Cable damage due to transport or laying are usually discovered by sheath testing. Therefore, on-site PD measurements on cable systems can focus on cable accessories. Different kinds of unconventional PD coupling methods based on e.g. capacitive or inductive sensors led to increased PD sensitivity at the accessories (joints) compared to conventional PD detection at the cable end. However, these types of sensors have to be implemented during the laying of the cable, since retrofitting of direct buried cable system is often not possible. To avoid all these drawbacks, PD sensors have to be placed in easily accessible and noncritical part of the HV cable system. A promising approach is the use of inductive PD sensors at cross bonding links of long HV cable systems [34]. It is recommended to carry out PD measurements synchronously at all accessories on a per phase basis. Selective PD measurements need a potential-free connection from the accessories to the storage and visualization unit; this can be achieved using optical fibers. TEST SET-UP Remote-controlled PD acquisition units (AUs) are positioned at all accessories together with a quadrupole and a power supply. The sensitivity and signal-to-noise ratio of the PD measurement can be optimized by employing fully digital signal processing capabilities; the PD signal is sampled over a wide frequency range and the measurement is done in a region of the spectrum with a low noise level, by choosing an appropriate center
Page 57
Guidelines for PD Detection
frequency and bandwidth. The PD data is transferred via optical fiber to the central storage and visualization unit (PC), and the PD detection units are controlled (Figure 60).
FIGURE 60: SETUP OF MULTI-CHANNEL PD MEASUREMENT SYSTEM (OPTICAL COMMUNICATION)
MEASURING PROCEDURE An example of a three step verification procedure of a PD system performed during after-installation testing of a 20 km long, 400 kV XLPE cable system is described. The measurements were synchronously performed on all accessories, twenty joints and two terminations [35].
Firstly, the internal test signal generators of each PD measuring unit are activated to verify full functional readiness of the system. In the second step, calibration pulses of 500 pC are fed in via one of the cable outdoor terminations. The calibrator operates synchronously with the PD measuring units via optical fiber bus. To trace these calibration pulses along the 20 km cable route, it became necessary to average measurement data from the capacitive PD sensors. The 500 pC calibration signal and its part-reflection at the next joint can clearly be distinguished from the background noise level even over a distance of 20 km (Figure 61).
Page 58
Guidelines for PD Detection
FIGURE 61: PROPAGATION OF CALIBRATION PULSES ALONG 20 km XLPE CABLE LINE [35]
FIGURE 62: PROPAGATION OF CALIBRATION PULSES ALONG 20 km XLPE CABLE LINE
The calibration pulses detectable at all accessories were transformed into the frequency domain using the FFT algorithm. A partial result is shown in Figure 63, where the severe attenuation of the higher-frequency signal components with increasing distance from the feed location of the calibration signals can clearly be discerned. The drop in the spectra at approx. 4.5 MHz can be attributed to the specific resonance conditions at the location where the calibration pulse was injected.
Page 59
Guidelines for PD Detection
FIGURE 63: SPECTRA OF CALIBRATION PULSES
Prior to each voltage test for one phase, a “wire test” can be conducted which involved measuring the PDs produced by a copper wire connected to high voltage and directed towards ground, in order to verify the functional effectiveness of all PD acquisition units installed and of the processing computer. The frequencydependent attenuation of these PD signals can be highly advantageous for demonstrating the selectivity of PD measurements at the individual accessories. The center frequency and bandwidth of the PD measurement was such that interference from the test setup played practically no role at all. Figure 64 depicts this influence of the measuring frequency. The left-hand side of the diagram shows the PD patterns of all AUs at 6 MHz midfrequency (exception: PD station 1.24 – also the location of the SF6 sealing end – the mid-frequency is 1 MHz). PD stations 1.1, 1.2 and 1.3 were measured with one capacitive and two inductive PD sensors near the supply side (SF6 sealing end), and exhibit similar PD amplitudes to the wire test. PD stations 1.4, 1.5, 1.6 and 1.7 at the first four joints exhibit a steep decrease in PD amplitude with increasing distance from the supply side. PD station 1.8 (5th joint), measured only uniform background noise. In the right-hand half of Figure 64, the mid-frequency was lowered to 600 kHz. With respect to the spectra shown in Figure 63, the distance-dependent attenuation of the PD signals was reduced far enough to ensure that the PD patterns of the wire test at all cable accessories could still be clearly detected over the basic background noise level. At this measurement frequency, the four phase-locked interference pulses caused by the inverters of the resonance system also emerged quite prominently in all PD patterns.
FIGURE 64: PD PATTERNS OF WIRE TEST
Page 60
Guidelines for PD Detection
CONCLUSIONS Because the attenuation of external noise increases proportionally with the cable length, the PD noise floor measured at cable accessories is similar to (and sometimes even lower than) that of shielded laboratories. In this context, synchronous PD measurements allow early detection of PD faults and can be used to localize a PD fault to a specific cable accessory.
Even for the supply side (cable sealing end), where external interference signals enter without any attenuation and selectivity could not previously be assured with any certainty, synchronous multi-channel PD measurements open up new, significantly improved options for unambiguous PD assessment.
3.7
Case studies
3.7.1
Case Study 1: Maintenance test of a 66 kV power cable with single-ended PD detection and PD localisation Maintenance testing by means of DAC was performed on a service-aged, 35-year old, 2.2 km long 66 kV XLPE insulated underground circuit, see Figure 65 and Figure 66 [32]. Starting from 1.1 Uo, PD activity up to 100 pC has been registered in one of the joints. Increasing the test voltage up to 1.5 Uo resulted in concentrated PD activity appearing in three joints. Based on this test, it has been concluded that this cable section can be energized for network operation with a possible risk of a failure during operation. Due to the fact that PDIV was very close to Uo and increased network stresses may result in an inception and increase of PD activity, the risk of a failure depends on over-voltage stresses which may occur during operation. Replacement of the joint was recommended. However, if this wasn’t done, it was recommended to perform the next maintenance tests within approximately 6 months in order to evaluate the progress of degradation at the above mentioned locations by comparing the change of PD activity over time. 80
80
60
60 40 Voltage (kV)
Voltage (kV)
40 20 0 -20 0
1
2
3
4
5
6
7
-40
8 9 Time (ms)
10
11
12
13
14
15
16
20 0 -20 0
2
3
4
5
6
7
8 9 Time (ms)
10
11
12
13
14
15
16
1
2
3
4
5
6
7
8 9 Time (ms)
10
11
12
13
14
15
16
-60
-80
-80
2'500
2'500
2'000
2'000
1'500
1'500
PD (pC)
PD (pC)
1
-40
-60
1'000
1'000 500
500
0
0 0
1
2
3
4
5
6
7
8 9 Time (ms)
10
11
12
13
14
15
0
16
L1
L2 80 60
Voltage (kV)
40 20 0 -20 0
1
2
3
4
5
6
7
8 9 Time (ms)
10
11
12
13
14
15
16
1
2
3
4
5
6
7
8 9 Time (ms)
10
11
12
13
14
15
16
-40 -60 -80
2'500
PD (pC)
2'000 1'500 1'000 500 0 0
L3 FIGURE 65: PD PATTERNS OBSERVED AT 1.5X UO DURING MAINTENANCE TESTING OF A 30-YEAR CABLE
(30 year old, 2.2 km long, 66 kV XLPE insulation cable installed underground)
Page 61
Guidelines for PD Detection
FIGURE 66: PD MAPPING AS MADE UP TO 1.5X UO DURING MAINTENANCE TESTING OF A 30-YEAR OLD CABLE
(30 year old, 2.2 km long, 66 kV XLPE insulation cable installed underground)
3.7.2
Case Study 2: After-laying test of a 10 kV power cable with single-ended PD detection and PD localisation A newly installed, 2.1 km long, 10 kV XLPE insulated underground cable circuit has been tested in accordance to IEC 60502 which recommends voltage withstand testing using sinusoidal DAC voltage up to 2xUo. It has been decided to perform monitored withstand testing at 2.0xUo. Standardized PD detection has been applied for the duration of the test, and no breakdown has been observed. It follows from Figure 67 and Figure 68 that internal PD activity has been registered in a joint in phase L3. The joint has been investigated and the PD source (in a crimping tube) has been identified. After the repair the complete cable system was PD–free and the test has been considered successful.
FIGURE 67: PD RESULTS OF AN AFTER-LAYING TESTING OF A 10 kV 2.1 km LONG XLPE CABLE SECTION
Page 62
Guidelines for PD Detection
FIGURE 68: INVESTIGATION OF THE JOINT AT 955 m HAVING PD UP TO 800 pC AT 2xUO
3.7.3 Case Study 3: Double-ended PD measurement on medium voltage cable This example shows a measurement on a 644 m long medium voltage cable with PD detectors at both ends of the cable. The PD detector A is on the left end of the cable and its output is plotted in black Figure 69. PD detector B is placed at the right end and its output is plotted in grey. The PD detectors are synchronized with a time resolution of 10 ns. This results in an accuracy of about 2 m for locating the PD origin. As can be seen in Figure 69, the PD pulse is reaching the PD detector A before detector B. That means the PD location is nearer to the left end of the cable than to the right end. The time delay between the two pulses is about 0.25 µs, which corresponds to a distance of approximately 45 m of travelling length. It can be concluded that the PD origin must be 45 m from the middle of the cable, e.g. at 277 m, seen from detector A.
FIGURE 69: DOUBLE SIDED MEASUREMENT ON A 644 m LONG MEDIUM VOLTAGE CABLE
Page 63
Guidelines for PD Detection
3.7.4
Case Study 4: After-laying test 220 kV cable with single-ended PD measurements at both ends A newly installed, 13.3 km long, 220 kV XLPE insulated underground cable circuit has been tested, applying 1.3xUo, see Figure 70 - Figure 72 [32]. It has been decided to perform a PD monitored DAC withstand test. PD activity has been observed in phase L1 as the test voltage was increased, starting from 0.2Uo. An increase in the test voltage has resulted in an increase of PD activity and at 0.4xUo test voltage, a breakdown at the discharge site occurred. Using PD mapping, the PD concentration at 5.3 km has indicated the breakdown position in the cable. As a result, the after-laying testing using damped AC voltage proved to be effective for monitored testing of a 13.3 km long length of newly installed 220 kV cable. The defect produced PD before an actual breakdown occurred, and with TDR analysis the PD defect location could be determined. The other 2 phases have fulfilled the after laying conditions and successfully passed the test. No internal PD activity in the cable insulation and accessories and no breakdown occurred during the complete test. The measurement was repeated from the other side of the cable. Again the PD activity occurring before the breakdown could be localized at 8 km, which is the same location seen from the other side (13.3 km – 8 km = 5.3 km).
FIGURE 70: ON-SITE TESTING OF A 220 kV 13.3 km LONG XLPE CABLE CIRCUIT.
The coupling capacitor is connected to one of the cable section phases. 60
40
40
20 0 -2 0
0
1
2
3
4
5
6
7
8
9
-4 0
10 11 Ti me (m s )
12
13
14
15
16
17
18
19
20
21
Vol tag e (k V)
Vo tl a g e ( kV)
p 60
q
y
20 0 - 20
0
1
2
3
4
5
6
7
8
9
10 T m i e ( ms )
11
12
13
14
15
16
17
18
19
0
1
2
3
4
5
6
7
8
9
10 T m i e ( ms )
11
12
13
14
15
16
17
18
19
- 40
-6 0
- 60
6 00
600
5 00
500
4 00
400
P D (p C)
PD ( pC )
;
3 00 2 00
300 200 100
1 00 0
0
0
1
2
3
4
5
6
7
8
9
10 11 Ti me (m s )
12
13
14
15
16
17
18
19
20
21
b
a
FIGURE 71: PD PATTERNS AS OBSERVED DURING THE VOLTAGE WITHSTAND TESTING OF A 220 kV XLPE CABLE
13.3 km long 220 kV XLPE underground circuit: a) example of PD pattern at 0.2x U0 of phase L1, b) example of PD pattern at breakdown voltage of 0.4x U0 of phase L1
Page 64
Guidelines for PD Detection 600
380 550
360 340
500
320 300
450
280 400
260 240 PD (pC)
PD (pC)
350 300 250
220 200 180 160 140
200
120 100
150
80 100
60 40
50
20 0
0 0
1'000
2'000
3'000
4'000
5'000
6'000 7'000 Location (m)
8'000
9'000
10'000
11'000
12'000
13'000
0
1'000
2'000
3'000
4'000
5'000
6'000 7'000 Location (m)
8'000
9'000
10'000
11'000
12'000
13'000
FIGURE 72: PD MAPPING MADE UP TO 1.3X U0 DURING ON-SITE TESTING OF A 220 kV 13.3 km LONG CABLE CIRCUIT.
The PD concentration at 5.2 km distance indicates the breakdown site of phase L1 (left). Measurement from the other end confirmed this location at 8.1 km, as seen from the other end (right)
3.7.5 Case Study 5: Continuous PD monitoring experience in a MV cable system A continuous PD monitoring system was used for a new 45 kV cable system of 4136 m length (see Figure 73 a), with four cross bonding splices and two terminations [36]. The PD monitoring system was comprised of a HFCT PD sensor placed on the cable sheath of each cable accessory (splice or termination), a PD measuring instrument (MS) placed close to each set of three accessories, and a control acquisition system (CAS) to synchronize the acquisition of the six MS (S0, S1, S2, S3, S4 and S5). The CAS collects the synchronized raw data sent by each MS, removes the noise and performs the signal processing to locate PD sources and to analyse the evolution of the PD sources detected.
S0
FIGURE 73: DIAGRAM AND SIGNALS FROM CONTINUOUS PD MONITORING OF A 45 KV CABLE SYSTEM
Fig. 73a) Configuration of the continuous PD monitoring system using HFCT sensors placed at each cable splice, Fig. 73b) Raw data (green wave-shapes) of PD signal acquired by S0, S1, S2 and S3 sensors along with filtered PD signals (red traces), Fig. 73c) Arrival times of the PD pulses after applying the filtering tool, caused by an internal defect in phase R of the CB1 splice, detected by HFCT sensors: S0, S1, S2 and S3.
Page 65
Guidelines for PD Detection
Thirty-five days after energizing, the sensors S0, S1, S2, and S3 of the R phase detected the presence of a PD source. After filtering, PD pulses were distinguished from background noise (Figure 73-b). The time delay of the arrival times of the filtered PD pulses recorded by the four sensors S0, S1, S2 and S3 (Figure 73-c) allowed the location of the PD source to be determined to be 1061 m from the GIS-1, in the CB1 splice of phase R (Figure 74). The PD pulse frequency spectra of the PD pulses recorded by the four HFCT sensors are shown in Figure 75. High frequency components of the PD signal are lost when the PD pulse travels along the cable system. In the test described here, the upper cut-off frequency is not greater than 4 MHz when the PD pulse travels 1732 m (2792-1060=1732 m) along the cable system.
FIGURE 74: PD MAPPING OF THE CABLE SYSTEM MONITORED
A PD source located 1061, 35 m from the GIS-1, PD pulse rate indicated by the red bars, PD amplitude expressed in mV indicated by blue dots
FIGURE 75: FREQUENCY SPECTRA OF PD AT EACH MONITORING POINT OF THE CABLE SYSTEM
Figure 76 shows the phase-resolved PD pattern associated to the internal insulation defect detected. This PD source was active for 45 days until the splice was repaired. b)
a)
FIGURE 76: DETAILS OF THE PD SOURCE LOCATED IN SPLICE CB1
Page 66
c)
Guidelines for PD Detection Fig. 76a) Evolution of the PD source located in splice CB1, Fig. 76b) phase-resolved PD pattern of the PD source located in CB1, Fig. 76c) degradation caused by the internal defect
The evolution of the PD detected shows a decreasing of the PD amplitude (red curve of Figure 76-a) and an increasing of the PD pulse rate (green curve of Figure 76-a, expressed as the number of PD pulses detected per each grid period of 20 ms). When the joint was energized, no PD signals appeared, but after 35 days the first PD activity was detected. Figure 76-c shows the dissection of the splice. Clear evidence of insulation degradation due to PD activity was visible. An incorrect assembly of the splice was the reason for this defect. 3.7.6
Case Study 6: Comparison of insulation defect at a 66 kV GIS cable joint detected using HFCT and HF antenna A dual-circuit 66 kV GIS is interconnected to a power transformer via XLPE cables which contain a cable joint. These cables are aged more than 30 years. In the joint part, there was only one joint in which a pulse signal was detected via HFCT (100 kHz to 20 MHz) attached to the grounding cable. This cable joint was focused on to investigate the suitability of employing an available horn antenna (frequency range 750 MHz to 5 GHz) to detect the PD pulse signals in the HF range, as well as to examine the relationship between the signals picked up using the HFCT and the horn antenna. In this experiment, two HF antennas of the same type were placed in different positions P1 and P2 as shown in Figure 77: One antenna was placed facing the cable joint from 0.6 m away (Position P1) to detect the signals, while the other one was placed 2.2 away, facing in the opposite direction (Position P2) in order to detect ambient background noise, in order to compare it with the actual PD signal. The relationship between the signals and noise detected by the HF antennas, as well as between the signals simultaneously detected by the HF antenna and HFCT are discussed below.
(a) full view
(b) antenna position 1 (P1)
(c) antenna position 2 (P2)
FIGURE 77: PHOTOGRAPH SHOWING PD MEASUREMENT VIA HFCT AND TWO HF HORN ANTENNAS
The photograph of the measurement arrangement shows the HFCT and the two HF antennas for picking up the PD signals from the cable joint. Note that the punched metal panels were removed to allow passage of the HF signals.
Figure 78 shows a typical simultaneous measurement result of the two HF antennas and HFCT. In this measurement, the perforated metal panels enclosing the cable joint part were removed (Figure 77b).
Page 67
Guidelines for PD Detection
As shown in Figure 78, the HF antenna placed at P1, facing the cable joint, could measure the pulse signal together with the HFCT, while the HF antenna at P2, facing away from the cable joint, was not able to measure the pulse signal clearly.
FIGURE 78: TYPICAL SIMULTANEOUS MEASUREMENT RESULT OF THE TWO HF ANTENNAS AND HFCT
Figure 79 shows the relationship between a peak-to-peak amplitude Vpp measurement of the two HF antennas placed at the different position P1 and P2 along with the HFCT. As shown in Figure 79, the Vpp value measured by the HF antenna placed at P1 increased linearly with that measured by the HFCT, while the Vpp value measured at P2 did not show the same relationship to the HFCT. This demonstrates that the HF antenna at P1 measured the defect signal emanating from the cable joint part which is the same signal measured by the HFCT, while the HF antenna at P2 measured only the background environmental electromagnetic noise.
FIGURE 79: RELATIONSHIP BETWEEN SIGNALS MEASURED WITH HFCT AND 2 HORN ANTENNAS
The plot shows the relationship between the peak-to-peak amplitude Vpp of the two HF antennas, facing toward and away from the cable joints, respectively, and the HFCT. Note that if the punched metal walls were not removed, the HF antenna placed at P1 could measure the pulse signals even decreasing the amplitude by about 70% as shown in Figures 80 (a) and (b).
Page 68
Guidelines for PD Detection
(a) Measurement result
(b) Relationship between the HF antennas and HFCT
FIGURE 80: MEASUREMENT RESULTS WHEN PUNCHED PANELS WERE NOT REMOVED
3.7.7 Case Study 7: PD testing of short 220 kV XLPE cable system In-service PD measurements were performed on a 220 kV XLPE cable with a length of 100 m. The measurements were performed using HFCT sensors connected to the grounding shield of the cable at the GIS and at the OHL cable terminations. Multi-channel technique (3PARD) was used to allow a sensitive PD measurement. Figure 81a shows the three-phase synchronous PRPD patterns acquired at the GIS cable terminations. The 3PARD built from the acquired PD pulses from these three phases is also shown in Figure 81b.
a)
b)
c)
FIGURE 81: THREE-PHASE PRPD PATTERNS AND THEIR EQUIVALENT 3PARD DIAGRAM
The most relevant clusters were selected and their transformation back to the PRPD patterns was performed for individual analysis. Noise coming mostly from the overhead lines of the substation (located in the centre of the 3PARD diagram) was successfully separated out. The internal PD signal was detected at the middle phase of the cable line (S phase). The PRPD patterns of the signal after the separation and back transformation are presented in Figure 81c. As calibration of the set up was not possible in on-line conditions, the PD charge values displayed in the PRPD patterns are not valid. Page 69
Guidelines for PD Detection
To gain more information about the PD source type, the PD measurements were performed at various centre frequencies: 5 MHz, 7 MHz, 9 MHz, and 11 MHz (Figure 82). The frequency bandwidth of the acquisition unit digital filter was set to 3 MHz and was not changed during the measurements. It can be seen that the amplitude of the signal and the repetition rate of the PD pulses (about 130 pulses/sec) remained independent of the selected measuring frequency. As the attenuation of the PD signal is very low between 5 and 11 MHz, it can be assumed that PD activity takes place very close to the measuring sensor, namely, in the GIS cable termination of phase S.
FIGURE 82: PRPD PATTERNS OF THE PD SIGNAL AT DIFFERENT FREQUENCIES AFTER SEPARATION
After five weeks of operation, a failure occurred at the middle phase of the cable, where PD activity had been detected previously. The trace of the failure (Figure 83) was found 40 cm away from the edge of the termination stress cone, near the connecting clamp to the GIS bus bar.
FIGURE 83: FINDINGS AT THE GIS CABLE TERMINATION INSPECTION AFTER THE FAILURE
Page 70
Guidelines for PD Detection
3.8
Advantages and disadvantages of different methods
TABLE 6: COMPARISON OF CONVENTIONAL (IEC 60270) AND UNCONVENTIONAL PD DETECTION METHODS FOR CABLES
Conventional (IEC 60270)
Calibrated in pC
Unconventional HF measurement
HF/VHF measurement
(use of coupling capacitors and HF coupling devices) Calibrated readings in pC not possible
(use of HFCTs) Calibrated readings in pC not possible
Better applicability under on-site conditions
Applicable under on-site conditions, but strongly influenced by defect position, especially at VHF and UHF frequencies.
High sensitivity in low noise environment
Distributed measurements possible at both sides of the cable
Sensitivity influenced by the measurement frequency and the type and position of the sensors
PRPD reference patterns are available for different PD defects
PRPD patterns similar to conventional measurements
PRPD patterns similar to conventional measurements
Applicable off-line, not on-line
Applicable off-line, not on-line
Applicable both off-line and on-line
Not possible to locate PD
With time domain reflectometry, PD localization possible. For long cable lengths, measurements on both ends of the cable recommended
PD localization possible on the cable accessories, multiple sensors on various locations are needed
Standardized procedures for type testing, factory testing and on-site testing
No standardized procedure for onsite commissioning, assessment, or monitoring
No standardized procedure for onsite commissioning, assessment, or monitoring. Comparison of different detection systems not possible
Difficult to suppress EM interference on-site
Page 71
Guidelines for PD Detection
3.9
Summary and important aspects
1. When performing PD detection on power cables, several aspects need to be considered: a. the length of the power cable and the PD signal attenuation characteristics e.g. for a few tens of meters, below some few hundreds of meters (no joints), below a few kilometers, below 20 kilometers, and longer b. the applied PD detection and localization methodology e.g. TDR to detect PD in the cable insulation and cable accessories, distributed HF sensors to detect PD in accessories only c. cable construction and the type of accessories e.g. straight joints, cross-bonding joints, open-air terminations or GIS/transformer connectors d. knowledge about the sensitivity and frequency response of sensors connected to the cable circuit, as well as the possibility of carrying out a performance check on-site. 2. When making PD measurements using unconventional methods, it is not possible to quantitatively correlate or calibrate the received signal level in terms of PD charge (pC or nC). 3. When applying the described unconventional PD detection methods, the parameters under 1a…1c should be taken into consideration. 4. Using appropriate signal processing techniques, e.g. when selecting frequency ranges to obtain the best possible SNR; one should consider the overall detectability of the PD process in question, i.e. selecting frequency ranges within which the actual PD type can be measured. 5. When performing unconventional PD measurements, verification checks should be made whenever possible, even during on-site measurements, to assure the operator that the measurement system is performing in the way expected. Regarding PD detection on power cables, this study has provided the following two common learning points: 1. On-site testing of power cables is generally carried out using energizing methods such as AC, DAC, and VLF. The advantages and disadvantages of the two major PD detection methods, conventional and unconventional, are summarized in Table 6. 2. There is a large number of parameters differentiating which PD detection system shall be used as well as specifics of the individual power cables under the test. The key aspects which need to be considered have been defined in section 3.9 As a result, for on-site PD testing of power cables, several gaps are still present which require further attention and investigation. In particular, from a practical point of view, there is a need to work out uniform and simple calibration-, sensitivity-, as well as performance-check procedures which can be applied independently to both the test voltage shape and the PD measuring system applied. Based on such simple procedures and feedback from users, better defined and more specific guidelines can be subsequently worked out.
Page 72
Guidelines for PD Detection
4 POWER TRANSFORMERS 4.1
Types of partial discharges
Partial discharges in the insulation system of transformers can occur in a specific location where the electrical field stress exceeds the maximum withstand level of an insulation material. In this region, a continuous breakdown occurs at nominal voltage (or below it), or at test voltage level during the acceptance test in the factory or on site. PDs in new transformers occur due to flaws, oil contamination, or deviation from permissible tolerances in the manufacturing process. Another possibility is that tests applied in advance to the test object triggered some kind of damage which results in PD signals appearing during the PD tests. For transformers in service, electrical, mechanical, thermal, and/or chemical stresses, alone or in combination, may cause PD to occur. In most cases such PD activity can be detected by means of analysis of dissolved gases in the insulating oil, known as 'dissolved gas analysis’ (DGA). The type of PD-source in the insulating system of the transformer can be estimated by analysis of phaseresolved PD patterns which are recorded during the PD measurement. The phase-resolved PD pattern is reflecting the physical phenomena (statistical behavior) of the specific PDsources. For PD to start and continue, a starting electron is necessary for generating PD in a specific localized region as defined above. The availability of starting electrons is strongly dependent on the PD source itself, the insulation material (solid, fluid or gas) and on the position of the PD source with respect to the electrodes, which defines the electrical field strength. Based on the physics of electric discharge processes in the weak region, and using the statistical evaluation processes for generating the PRPD pattern according to [37], five typical PD-patterns can be defined if the sensitivity of the measuring circuit is sufficient to acquire the signal. By investigating the same PD-source using different measurement principles such as e.g. electrical (according to IEC 60270), UHF or acoustic methods, theoretically a similar PRPD pattern can be obtained if the propagation times and the attenuation of the different measurement principles are normalized. Unfortunately, due to noise, simultaneous occurrence of different PD sources (superimposition), and changing amplitudes of the different PD sources present, in reality the five typical PD patterns appear in many variations. Consequently, there are only a few PD patterns that exhibit constant behavior during the test. Interpretation and screening of the correct type of PD pattern from the real measured PD pattern requires considerable experience along with a good deal of interpolation / abstraction capability. An overview of the typical PD sources in the transformer insulating system together with their typical PD pattern and their typical behavior during the test is presented in [37]. If there is a clear indication of internal PD activity in the transformer insulating system, localization of the PD source must follow to be able to make an accurate assessment; the plausibility of the type of PD and its location in the transformer must also be checked, based on the transformer design documentation. 4.1.1 Pulse shape and frequency content High precision measurement of PD current pulse waveforms is important to underpin insulation diagnosis techniques using the UHF method, because it gives us fundamental information about the signal source [38]. So far, G.P. Cleary, et al [39] constructed a system for the measurement of PD current pulse waveform in insulation oil which had a frequency response up to 3 GHz. They showed that the shortest rise time of PD pulses measured in mineral oil was 2 ns for the positive pulse and 0.9 ns for the negative one. The performance of the measurement system, especially the sampling rates and bandwidths of digital sampling oscilloscopes, has drastically improved in the meantime. Indeed, using such an improved instrument, S. Ohtsuka, et. al. [38] have
Page 73
Guidelines for PD Detection
constructed a sophisticated measurement system by expanding the frequency bandwidth up to the SHF band for the SHF_PDPW (Super High Frequency wideband PD current Pulse Waveform measurement) system. They have investigated PD current pulse waveforms in SF6 gas [38], SF6 substitute gases [40], and insulation oil [41], [42] with this system. Their results showed that the negative PD current pulse waveforms in SF6 gas and insulation oil exhibit very steep rise times in the range of a few tens of picoseconds and also showing a clear difference between the applied voltage polarities of the PD current pulse waveforms. In this section, PD current pulse waveforms measured with the SHF_PDPW system in mineral oil were introduced by taking into consideration the effects of the degassing treatment of the mineral oil. Also, two terms, fBW and fBWpro – pro-related to the frequency bandwidth - were used to indicate the influences of frequency bandwidth on the PD current pulse waveforms. The fBW term is the frequency bandwidth of the oscilloscope used for the measurement. The fBWpro term is the processed frequency bandwidth by the low pass filtering (LPF) processing with the Butterworth filter (filter order N=4). Two degassing treatments for the mineral oil poured into the PD test chamber were performed [42]:
For more than 6 hours by means of a vacuum pump, then backfilling with N2 gas at 0.1 MPa, followed by a stability period of more than 12 hours. This conditioning is defined as good degassing treatment,’Good_DT’. For app. 5 minutes by vacuum pump. This conditioning is defined as poor degassing treatment, ‘Poor_DT’. There was a small void visible in the oil for the Poor DT. This condition is defined as ‘Void_DT’.
The positive and negative partial discharge inception voltage (PDIV) depended on the degassing treatment condition [5]. Figure 84 (a) and (b) show typical examples of positive and negative PD current pulse waveforms with respect to the different degassing treatment conditions (1) to (4). At positive polarity, a number of PDs were generated forming a cluster which maintains itself for a few µs for the Good_DT (Figure 84 (a-1)) and the Poor_DT_1 (Figure 84 (a-2)) [43]. On the other hand, single pulse waveforms in the sub ns range were observed in the negative polarity, irrespective of the degassing condition (Figure 84(b)). In this figure (b), the rise times of Figure 84 (b-1) to (b-4) were 15.2 ps, 13.2 ps, 151 ps, and 173 ps, respectively. However, the pulse waveform was observed even in the positive polarity of the Void_DT (Figure 84 (4)). In Figure 84 (a-3) and (a4), the rise times were 247 ps and 140 ps, respectively. Additionally, in the case of the Poor_DT, two types of waveforms were observed: Poor_DT_1 and Poor_DT_2 as shown in Figure 84 (2) and (3). Regardless of the polarity, these waveforms Poor_DT_1 and Poor_PDT_2 were similar to the Good_DT and the Void_DT waveforms, respectively. When we processed the waveforms of Figure 84 (b-1) to (b-4) using the LPF (low-pass filtering) processing of fBWpro=1 GHz, the current peak Ip decreased and both the pulse width and the rise time (tr) increased; consequently the rise times of the processed waveforms (1) to (4) became 284 ps, 620 ps, 363 ps, and 350 ps, respectively. As seen in the processed waveforms, no significant differences in the waveforms between the degassing treatment conditions were observed. Namely, when PD current pulse waveforms were measured with the conventional measurement system of fBW=1 GHz, differences in PD current waveforms of the three conditions could hardly be observed. According to the discussion on the fBWpro dependences, the PD current waveforms of the Poor_DT_2 and the Void_DT seem to have lower frequency content below 4 GHz, while the Good_DT and the Poor_DT_1 seem to have much wider frequency content, extending well over 32 GHz.
Page 74
Guidelines for PD Detection
(a) Positive (Va=24kV)
(b) Negative (Va=23kV)
FIGURE 84: TYPICAL POSITIVE AND NEGATIVE PD CURRENT PULSE WAVEFORMS FOR EACH DEGASSING TREATMENT
Measurements of the different conditions of the mineral oil were done with the SHF_PDPW system [42]
4.2
Electrical PD measurement according to IEC 60270
Circulating PD current impulses generated by external PD sources (e.g. in the test circuit) or by an internal PD source (in the insulation system of the transformer) are typically measured during factory acceptance tests by using an external coupling capacitor. For measurement of PD in transformers, the coupling capacitor can be replaced by using the measurement tap of the high voltage bushings. For this reason a typical set up is shown in Figure 85.
Page 75
Guidelines for PD Detection
FIGURE 85: TEST CIRCUIT FOR MEASUREMENT AT A TAPPING OF A BUSHING [2]
The bushing capacitance Ck acts as coupling capacitance which is connected parallel to the test capacitance Ca of the transformer. For power transformers, the measuring impedance (CD) is usually connected to the measuring tap of the bushing, whereby the capacitance Cm is connected in parallel. For bushings without a capacitive tap, an external coupling capacitance Ck must be connected in parallel with the bushing. According to IEC 60270 and IEC 60076-3, it is preferred to measure the apparent charge by using a defined wideband or narrowband measurement system. The PD measurement system is connected via a coaxial cable to the measuring impedance CD. The apparent charge q, measured in pC, corresponds to the charge transferred during the voltage drop between the parallel connected capacitances Ck and Ca. The voltage drop (and the corresponding charge transfer) can be triggered by an external event or an internal PD source. If PD activity is detected during the measurement, the PD source must be investigated. To assign correctly the right charge value to a given voltage drop, a calibration of the circuit must be done before a measurement is taken. For this purpose a calibrator is connected between the high voltage terminal and the ground of the bushing and a reference charge is injected. The PD measurement system is then adjusted accordingly to this reference signal. According to IEC 60076-3, the PD measurement shall be carried out in conjunction with the induced voltage test. Any wideband or narrowband measurement system which is designed and tested according to IEC 60270 and IEC 60076-3 may be used for this measurement. The first PD measurement should be made at a low test voltage level (e.g. 10 % of the rated voltage) in order to determine the background noise level. The maximum background noise level should be less than 50 % of the maximum apparent charge value specified for the transformer under test. The PD level must be checked at all high voltage bushings on the transformer. The best way to accomplish this is to apply a multi-simultaneous measurement, in which the PD measurement system is set up so as to detect PD activity at all bushings simultaneously.
Page 76
Guidelines for PD Detection
The PD test is considered passed if there is no continuous PD activity higher than the maximum admissible value and if there is no rising trend in the apparent charge amplitude during the long duration test. For further information, the reader is advised to use IEC 60076-3, IEC C57.12.90, or standards (e.g. IEEE) relevant for application to PD testing of power transformers. If a PD activity is detected, a deeper investigation of the PD event shall be performed. A PRPD pattern can be generated by any phase-resolved PD system. A PRPD analysis delivers important information for the identification of the type of PD source for following reasons:
The PD pattern helps to identify a type of PD source. Individual PD patterns are not influenced by the signal transfer function of the extended insulating system. PD patterns can be used to distinguish between superimposed PD defects on the basis of different statistical behavior.
FIGURE 86: TYPICAL EXAMPLE FOR A PRPD PATTERN GENERATED IN A PD MEASURING SYSTEM
4.2.1 Signal attenuation A well-known limitation of the electrical PD measurement is that the actual discharge cannot be measured directly at the defect site, and therefore the actual charge level of the PD remains unknown. A calibration can only include the signal path between the signal recording device and the terminals of the transformer. PD signals which originate farther from the terminal (bushing), i.e. within the transformer, are influenced by their propagation path. This path cannot be calibrated and the reading (apparent charge) of the PD measurement system will not be the same for the same defect occurring at different positions inside the transformer. The influence of changes in the location of the PD source on the resulting electrical signals at the transformer terminal was investigated using a laboratory setup. A cylindrical steel tank is equipped with a winding and an artificial PD source whose position (height) within the tank is adjustable. The artificial PD source is connected to ground on one side and consists of two copper plates connected through a capacitor and a gas-filled discharge tube (GDT). The latter is used because it provides a reproducible, phase stable, and constant charge amplitude source. The PD source is fixed on a threaded rod which can be adjusted in height along the winding [44].
Page 77
Guidelines for PD Detection
FIGURE 87: EXPERIMENTAL SET UP SHOWING MOVEABLE PD SIGNAL SOURCE
Using this experimental setup, a broadband electrical PD measurement is performed. The upper frequency limit of the quadrupole used is about 15 MHz. Its output time domain signal is converted into the amplitude density spectrum in frequency domain using the Fast Fourier Transformation. Figure 88 shows the amplitude density spectrum for four different positions of the artificial PD source along the winding, as measured at the upper (terminal) end of the winding. -40
IEC broadband
IEC narrow band
increased frequency
-50
dBm
-60 -70 -80 -90 -100
2,5 cm 10 cm 40 cm 80 cm
-110 100 k
1M f / Hz
10 M
FIGURE 88: SIGNAL POWER DEPENDENCY ON LOCATION AND MEASURING FREQUENCY
Broadband IEC: Narrowband IEC: Increased frequency range:
250 1 4,5
Page 78
300 30 1
Guidelines for PD Detection
An increase in signal damping due to the low-pass filter effect of the winding with increasing insertion depth of the PD source can be clearly recognized. Three distinct frequency ranges are highlighted in the plot, two of which correspond to the IEC 60270 mandated wideband and narrowband modes, respectively. The third range, at higher frequency and bandwidth, does not conform to IEC 60270, but is often used for on-site PD measurements in order to suppress external interference. Especially for high insertion depths of the PD source (e.g. the green trace for the 80 cm position), measurement at higher frequency (fm = 4.5 MHz, Δf = 1 MHz) hardly shows a measurable PD signal level. This effect can be even more pronounced for larger windings. In Figure 89 the apparent charge is measured in two frequency ranges to show the damping effect on the PD position along the winding. The PD measurement system is re-calibrated after each change of the frequency range.
FIGURE 89: DEPENDENCY OF MEASURED APPARENT CHARGE ON LOCATION AND MEASURING FREQUENCY
Measured according to IEC 60270; left side: IEC broad band, fc = 250 kHz, Δf = 300 kHz; right side: IEC narrow band: fc = 1 MHz, Δf = 30 kHz
Despite calibration and a reproducible, stable PD source, significantly different PD levels are measured even at low insertion depths due to the filtering effect of the winding. Using the wide-band frequency range, a strong decrease in the level at 60 cm and at 80 cm can be seen, which could be initially assumed to be a measurement error, but is confirmed by an increased number of measurement points at similar insertion depth [44].
4.3
UHF PD-measurement
UHF measurement is based on the fact that PD take the form of fast rise-time electrical impulses which radiate electromagnetic waves with frequencies up to the ultra-high frequency range (UHF: 300 MHz to 3000 MHz) in the surrounding oil. It has been established that picking up these UHF signals can be used as a trigger for acoustic PD localization and for onsite/online diagnostic PD measurements and is also suitable for on-line PD monitoring. The grounded transformer tank provides a high level of electromagnetic shielding against external disturbances at UHF frequencies (e.g. corona, mobile telephones, radar, etc.), plus UHF signals propagating inside the tank experience only moderate attenuation; therefore UHF methods achieve high sensitivity and are thus advantageous for on-site measurements, as seen in Figure 90 [45].
Page 79
Guidelines for PD Detection
FIGURE 90: PRINCIPLE OF UHF PD MEASUREMENT IN COMPARISON TO ELECTRICAL MEASUREMENT
Both the measured electrical and UHF PD levels both depend on the following:
The type and actual charge level of the PD source. The signal attenuation in the transmission path. The sensor sensitivity. The sensitivity of the measurement device.
4.3.1 Signal attenuation For the reliable detection of PD defects in large power transformers at UHF frequencies, it is essential that the UHF signals emitted by a PD can be measured everywhere in the transformer with little or no loss of UHF signal energy. For analyzing UHF signal propagation characteristics inside power transformers, a 210 MVA grid coupling transformer which had been intended for scrapping was instead prepared for attenuation experiments. The transformer was oil free but with its tank intact, which therefore acted as a faraday cage against external disturbances, as usual. More importantly, the transformer included all of its internal parts. In order to inject signals at several different locations to assess EM propagation through the transformer, 20 holes (Ø 5 mm) were drilled into the tank wall at various positions around the transformer tank. The locations of holes no. 15 – 20 are shown for the back of the transformer in Figure 91. The internal design consists of the three limb core with the winding and an on-load tap changer on the right hand-side. For emitting artificial UHF impulses, a 10 cm long monopole antenna was inserted through the holes and into the transformer tank. Using this antenna, a signal generator was used to inject pulse with a maximum amplitude of 60 V at 50 Ω. These UHF signals emitted inside the transformer were measured by a UHF probe installed at the oil filling valve. In order to achieve maximum sensitivity, the antenna was pushed fully into the tank (12 cm into the tank). The measurements were conducted with an oscilloscope of 3 GHz analogue bandwidth using a 300 MHz high pass filter. The diagram in Figure 91 left presents the maximum amplitude of UHF signals measured with the UHF probe, referenced to the position of the source (hole number where monopole was inserted), and additionally, the shortest distance between the source and probe. In Figure 91 right, the attenuation in terms of dB and the attenuation in terms of dB/m versus the position of the UHF source are shown. All measured values are thereby
Page 80
Guidelines for PD Detection
referenced to the highest measured amplitude (43.4 mV), here position 20, drilled behind the tap changer drive [46].
FIGURE 91: POSITION DEPENDENCE ON UHF SIGNAL ATTENUATION INSIDE A 210 MVA TRANSFORMER
Top: measurement positions on the side of 210 MVA transformer tank; bottom left: maximum amplitudes of UHF signals, bottom right: attenuation values of UHF signals inside transformer tank
Not all 20 positions are specified in the diagram. P1, P3, P5 and P17 were not considered, since the monopole was not able to be completely inserted into the tank. From the 14 positions measured, the absolute attenuation shows a wide variation, from 2 dB to 18 dB, whereas the distance-based attenuation exhibits less variation, between 0.5 dB/m and 2.5 dB/m. Overall, an average attenuation of 2.0 dB per meter can be assumed. Rough estimation of additional damping of the structure with help of the propagation paths shows that a signal which includes e.g. a complete winding in its propagation path is attenuated by around 6 dB. If the propagation path of the signal only touches/grazes the winding, then the attenuation is only half that value (3 dB). If the tap changer lies in the propagation path, an additional attenuation of 2 dB can be expected [46]. 4.3.2 Sensor sensitivity In Figure 92 two typical types of UHF sensors can be seen. On the left side a UHF drain valve sensor suitable for standardized oil valves DN50/DN80 is shown. It can be used for diagnosis measurement or retrofit of transformers with a UHF monitoring system. On the right side, a UHF plate sensor for direct mounting onto a transformer tank wall or manhole plates is shown. This sensor type can be installed in new transformers for UHF monitoring.
Page 81
Guidelines for PD Detection
FIGURE 92: EXAMPLES OF UHF PD SENSORS FOR HV TRANSFORMERS
Left: UHF PD drain valve sensor, suitable for retrofitting in standardized oil valves; right: UHF plate sensor, suitable for direct mounting into transformer tank wall (new transformers)
With electric PD measurement according to IEC 60270, the influence of the electric ’sensor’ (coupling capacitor and quadrupole) and the measurement device can be corrected using calibration. Although the actual PD level cannot be estimated [47], [43], [48] because of the unknown signal path attenuation and unknown ratio of internal capacitances, calibration allows the introduction of apparent charge as an acceptance level; this is why direct electric measurement is used for routine acceptance tests. A correction of sensor influence can also be achieved for the UHF method. To determine UHF sensor sensitivity, the UHF antenna factor (AF) must be known. The antenna sensitivity depends on its design and size in relation to the electromagnetic wavelength. Antennas are described by different characteristics, e.g. by the antenna gain or the antenna aperture. For antennas which are not defined by a physical area, such as monopoles or dipoles, the antenna gain is often specified by the effective length leff, or the antenna factor AF, which is defined for a receiving antenna as AF(f)=E(f)/U(f)
4.1
where U(f) is the voltage at the antenna terminals and E(f) is the electric field strength at the antenna. The effective length is defined by the inverse antenna factor. The smaller the antenna factor, the more sensitive the antenna, because a smaller electric field strength is required to generate the same terminal voltage. For the evaluation of the antenna sensitivity, a special setup is necessary with no external disturbances and no internal reflections of electromagnetic waves. Therefore a specially equipped EMC absorber room can be used, or a GTEM-cell. The AF shown in Figure 93 is measured by the use of a GTEM-cell for two realistic arrangements for a sensor in a DN50 tube of an oil filling valve and without oil valve. [49]. 80
without oil valve with DN50 oil valve
70
AF / dB/m
60 50 40 30 20 0
0.5
1
1.5 frequency / GHz
FIGURE 93: TYPICAL ANTENNA FACTOR OF A UHF PD SENSOR MEASURED IN GTEM CELL
Page 82
2
2.5
3
Guidelines for PD Detection Light Blue curve: UHF sensor direct mounted to the cell Dark Blue curve: UHF sensor installed via standard DN50 drain valve to the cell
Experience from various narrowband measurements confirms that the highest sensitivity within the UHF frequency band between 300 MHz and 3 GHz can be achieved at around 510 MHz, where the AF in Figure 93 is low. It is advantageous to shift the probe deeper into the transformer. The AF is then lower and the antenna is more sensitive to the EM waves emitted by PD. Determination of the gain in sensitivity is only valid for a specific frequency. It can be concluded that above 300 MHz, the increased sensitivity gained by shifting the probe deeper into the transformer is approx. 10 dB, or approx. a factor of three in amplitude for time domain signals. The reason for this behavior is the directivity characteristic of the probe. This can be investigated and demonstrated by high frequency field simulations as shown in Figure 94 [50].
FIGURE 94: DIRECTIVITY CHARACTERISTIC OF A TYPICAL CONE-SHAPED UHF PROBE SIMULATED AT 500 MHz
The simulated directivity is given in terms of dBi, i.e. the sensitivity is correlated to the isotropic radiator. In the green-colored region, the sensitivity of the probe head is equivalent to the sensitivity of the isotropic radiator, whereas the orange and red regions indicate a higher sensitivity of approx. 6 dBi. That translates to probe sensitivity four times higher for EM signals arriving from the side rather than from toward the face of the probe (meaning from above in Fig. 94). Therefore in order to get the most sensitive part of the probe inside the transformer, i.e. its side, the probe head itself must be inserted inside the tank [46].
FIGURE 95: DEPENDENCY OF ANTENNA FACTOR ON INSERTION DEPTH FOR UHF PD SENSOR IN OIL-FILLED GTEM CELL
Page 83
Guidelines for PD Detection
The insertion depth of the antenna has a decisive influence on the antenna factor and thus on the sensitivity as shown in Figure 95. The antenna is still inside the gate valve at Pos. 1, which is an undesirable position, but one which does occur when making practical measurements. Pos. 2 is the most common case for transformer installations. The antenna just reaches into the transformer’s tank volume. Further insertion is often not possible because sufficient distance to the HV windings must be maintained to ensure safe isolation. In this case, the antenna is outside the cell’s homogenous field. The other positions 3 to 5 are within the cell’s test volume. Best AFs are achieved at high insertion depth, see Pos. 5, but improvements are small at frequencies f > 1 GHz. The local maximum caused by the gate valve is shifted to lower frequencies. The worst case installation at Pos. 1 increases the AF in the entire frequency range, for example at the local maximum at 400 MHz. Additionally, new peaks at 1.9 GHz / 2.1 GHz and frequencies f > 2.6 GHz occur [51]. 4.3.3 Recommendation for standardized valves for retrofit of UHF sensors Most available UHF sensors are designed for standardized DN50 or DN80 gate valves. The sensors also fit to guillotine and ball valves. In some cases an adapter flange may be required. Other valve types without a straight-through opening, e.g. globe, butterfly, and diaphragm valves are not supported.
FIGURE 96: POTENTIAL VALVE TYPES FOR RETROFIT OF UHF SENSORS; VALVES WITH STRAIGHT-THROUGH OPENING/DUCT
FIGURE 97: OIL VALVES WITHOUT STRAIGHT OPENING/DUCT; RETROFIT OF VALVE-TYPE UHF SENSORS NOT POSSIBLE
Page 84
Guidelines for PD Detection
4.3.4 Recommendation for a dielectric window for installation of UHF sensors On new transformers, UHF sensors can be installed directly into the tank wall; such UHF plate sensors do not need oil valves for installation. They can be installed at a stainless steel flange with a dielectric window at the transformer tank wall according to Figure 98 [52].
FIGURE 98: UHF PLATE SENSOR WITH STAINLESS STEEL FLANGE AND DIELECTRIC WINDOW
FIGURE 99: EXAMPLE FOR DIMENSIONS OF STAINLESS STEEL FLANGE AND DIELECTRIC WINDOW
Following the specification for transformer tanks the dielectric window should fulfill the following requirements:
Resistance to mineral transformer oil and natural/synthetic ester. Vacuum stability of 0.15 mbar for a minimum holding time of 72 h. The leakage rate should not exceed 5 mbar/s after stop of the vacuum pump.
Page 85
Guidelines for PD Detection
Pressure: 5 bar. Temperature: 120 °C
4.3.5 Interpretation PD pattern analysis of UHF PD measurement is comparable to the methodology employed in electrical PD measurements, but there is no correlation of the received signal strength – i.e. the antenna output signal voltage or energy - to the apparent charge measured with the electrical method. Therefore in UHF PRPD patterns, the y-axis is usually scaled in UHF signal amplitude [dBm, dBuV, mV] or UHF signal energy [pJ]. Both values show the same qualitative correlations. In Figure 100 a typical example of a UHF PRPD is shown. 30
14
10
10
8 6
5
4
number of PD
amplitude / mV
12
2 50
100
150 200 250 phase / degree
300
350
1
FIGURE 100: TYPICAL EXAMPLE OF A TRANSFORMER PRPD PATTERN OBTAINED USING UHF TECHNIQUES
4.4
Acoustic PD-measurement
In addition to the measurable electrical signals described above, partial discharges in oil appear also generate acoustic impulses which propagate as waves in the ultra-sonic range (20-1000 MHz) [53]. 4.4.1 Signal attenuation For performing acoustic PD measurements on transformers, piezoelectric sensors are employed, mounted on the outside of the transformer tank. The velocity of propagation of acoustic signals in oil for operational temperatures between 50°C and 80°C may e.g. vary from around 1240 m/s to 1300 m/s [54]. Basically, insulation materials exhibit a low-pass character [55] and acoustic attenuation increases approximately as the square of the frequency f2 [56]. PD longitudinal/ transversal
transversal
oil
reflection
oil path steel path
.
tank sensor FIGURE 101: ILLUSTRATION OF THE STRUCTURE-BORNE PATH PROBLEM
As shown in Figure 101, because the acoustic sensor is not directly located normal to the PD source, the propagation path (simplified in the drawing as 'oil path', may possibly include e.g. pressboard and other winding components as well) and steel [57]. Page 86
Guidelines for PD Detection
Figure 101 illustrates the so-called structure-borne path problem which can cause erroneous early arrival times of the acoustic signal. The mechanical waves (vibrations) encountering the transformer housing create an alternative propagation path via the tank wall whose acoustic propagation velocity is much higher than that of oil. Calculations based on using the arrival times of signals which have travelled partly over structure-borne paths but assuming an average acoustic propagation velocity valid for oil for all signals will lead to incorrect estimates of the distance between the PD source and the sensor.
FIGURE 102: ACOUSTIC PD SIGNAL WITH KNOWN ARRIVAL TIME HIGHLIGHTED PLUS STRUCTURE-PATH COMPONENT’
Systematic investigations of this matter were carried out [58]. An acoustic sensor mounted on a steel plate was gradually moved from a position perpendicular to an acoustic source in oil to positions involving a growing proportion of the acoustic signal path via a metal plate (Figure 101). Depending on the angle Ψ, three regions can be distinguished where different wave types are predominantly stimulated. For PD location using an average sound velocity (propagation speed in oil), only signal portions of the direct path should be denoted as 'PD-signal'. The structure-borne component can thus be treated as 'interference’ instead (Figure 102) [59]. 4.4.2 PD localization Regarding localization of the PD source, two main approaches can be found: (i) on the one hand, alterations of the signal amplitude or deformations of the signals shape along the propagation path can provide hints for a source location, (ii) on the other hand, measured arrival times are used to calculate the origin of signals (often referred to as ‘triangulation’). S3 (x s3 , y s3 , z s3 )
S (x s4 , y s44, z s4 ) Si (x si , y si , z si )
Di
D3 D4 PD (x, y, z)
D2
S2 (x s2 , y s2 , z s2 )
D1
S1 (x s1 , y s1 , z s1 ) FIGURE 103: EXTERNAL ACOUSTIC SENSORS ON A TRANSFORMER TANK WITH A PD INSIDE USING CARTESIAN COORDINATES
Figure 103 shows a schematic view of a transformer tank to which i acoustic sensors are attached, a PD source inside, and with the resulting distances Di from the sensors Si to the PD source. Such arrangements form the geometric basis for the required mathematical formulations. The appropriate nonlinear observation equations are in the simplest case characterized by sphere functions which intersect at the PD origin.
Page 87
Guidelines for PD Detection
Depending on whether mixed-acoustic (i.e. triggering with the electric or electromagnetic PD signal) or allacoustic measurements are used, the number of unknowns is three (space coordinates (x, y, z) of the volume containing the PD source) or four (an additionally unknown temporal origin) respectively. Hence an exact spatial location of the PD source in the determined case requires at least three or four usable acoustic sensor signal arrival times. Often the SNR is too low in order to determine the arrival times with sufficient accuracy. In these cases, increasing the SNR by suppressing noise using signal averaging is usually a very effective method. This implies some prerequisites: (i) the signals are repetitive and it is possible to get a jitter-free trigger on them, (ii) signal and noise are uncorrelated and (iii) the noise is supposed to be white (white noise should feature a fairly constant spectral density in the frequency range investigated). During an averaging process, the noise included in the signals tends towards its statistic mean value which is zero, if the noise characteristic is indeed white. The repetitive part of the signal is superimposed constructively and remains unaffected. The theoretical maximum signal-noise-ratio gain is N0.5 where N is the number of superimpositions [59].
FIGURE 104: COMPARISON BETWEEN PURE ACOUSTIC AND UHF-TRIGGERED ACOUSTIC PD ACQUISITION [59]
The black trace is from a purely acoustic-based measurement of a single 132 pC PD pulse, while the orange trace is comprised of 500 superimpositions of acoustic signals with a maximum amplitude of only 9 pC, but obtained using UHF triggering; both signals come from the same experimental arrangement [59]
To be successful with acoustic averaging, a stable trigger of an actual signal related to the PD with a higher sensitivity/SNR than the acoustic signals is required. Such a combination of two PD signals of different type and sensitivity is classically used in test laboratories of transformer manufacturers, where the sensitive/higher SNR electric PD signal is combined with the acoustic signals to locate PD. The same holds for a UHF-acoustic combination, since comparative investigation on the sensitivity revealed a higher UHF sensitivity, especially for hidden defects. Figure 104 shows a comparison between an acoustic PD single pulse signal with an apparent charge of 132 pC and a UHF-triggered, averaged acoustic PD signal with 500 superpositions of maximum 9 pC. Although both signals were recorded within the same experimental arrangement (identical PD source, sensor and sensor position and an equal signal amplification), the single pure acoustic PD pulse acquisition showed no clearly observable information. However, the averaged acoustic PD signal with 500 superimpositions revealed an obviously visible pulse. The UHF PD signals which were used to trigger the averaging process were detectable for discharge levels all the way down to 5 pC apparent charge. Regarding the experimental arrangement, an attempt was made to model the acoustic propagation processes realistically. For that purpose, the configuration involved a disc-winding package at high voltage surrounded by two pressboard cylinders with a stimulated PD at the inner side, immersed in an oil-filled transformer tank.
Page 88
Guidelines for PD Detection
4.5
Case studies
4.5.1
Case Study 1: On-line PD measurement in a single-phase power transformer 220 kV/132 kV, 80 MVA The main challenges involved in on-line PD measurement are noise suppression, locating the PD source(s) within the transformer, and the pulse clustering of each PD source that can appear in the power transformer. In order to analyse the insulation condition of a single-phase power transformer 220 kV/132 kV, 80 MVA on-line PD measurements were performed [60], [61]. PD sensors were installed as follows:
At both bushings on the 220 kV and 132 kV sides, in order to acquire PD signals by means of coupling impedances. These coupling impedances were set up for each capacitive bushing in order to get a band-pass filter response (400 kHz-20 MHz) suitable for acquisition of PD signals (Figure 1-a & 1-b. Sensors PD_S1 & PD_S2). At the earth connection of the tank via an HFCT sensor with a band-pass filter response of 100 kHz to 20 MHz (see figure 1-c. Sensor PD_S3).
FIGURE 105: A) SENSOR PD_S1 IN THE 220 kV BUSHING, B) SENSOR PD_S2 IN THE 132 kV BUSHING, C) SENSOR PD_S3. HFCT PLACED AT THE TANK EARTH CONNECTION
A measuring system with a bandwidth of 300 kHz to 30 MHz using an automatic filtering noise suppression tool was applied. The noise suppression tool is developed on the basis of the wavelet transform plus additional statistical treatment. The effectiveness of the noise suppression using wavelet transform is clearly demonstrated by observation of the residual background noise after filtering. It is so low that high-definition phase-resolved PD patterns can be seen clearly above the base-line level (PRPD patterns in the first row of Table 7). Although the major portion of background noise was rejected by means of the wavelet transform, the PRPD pattern recorded is the overlapping of several PD sources. Consequently, correct insulation diagnosis cannot be made by simple visual inspection of the PRPD pattern. These situations are very typical for on-line PD measurements. In recent years the technique of clustering the PD signals based on their impulse wave-shape has been being refined and is very appropriate to separate the overlapped PD sources. A generic damped sinusoidal wave-shape model, defined by the equation below, was applied to each recorded current PD pulse i(t) . Three characteristic parameters (f, , and ) of each pulse’s wave-shape were used (oscillation frequency, f, and two time constants and associated to the PD pulse envelope) to perform the PD clustering (Table 7). i(t ) sin(2 f t )
(t t 0 )
e
Page 89
1 e ( t t 0 )
4.2
Guidelines for PD Detection
where t0 and parameters are the time and phase displacements for the instant t=0. In this PD record, three different PD clusters can be observed (Table 7):
Cluster #1, PD pattern related to an internal defect, which evidences the existence of an internal defect in the transformer insulation. Cluster #2, PD pattern unrelated to internal insulation defects (corona PD, surface PD signals). Cluster #3, pattern related to e.g. signals generated by power electronic systems. TABLE 7: RAW PD PATTERNS ACQUIRED BY MEANS OF THREE SENSORS (ROW 1); PD PATTERNS AFTER CLUSTERING PROCESSING (ROWS 2, 3, 4) Sensor PD_S1
Sensor PD_S2
Sensor PD_S3
f [7,7 MHz; 11,2 MHz] α [1,5 106 s-1; 77,4 106 s-1] β [2,4 106 s-1; 1,4 106 s-1 ]
f [7,4 MHz; 11,4 MHz] α [5,4 106 s-1; 64,1 106 s-1 m]
f [9,0 MHz; 11,4 MHz] α [1,9 106 s-1; 85,7 106 s-1] β [2,1 106 s-1; 90,2 106 s-1 m]
f [5,1 MHz; 6,4 MHz] α [0,5 106 s-1; 34,1 106 s-1] β [1,6 106 s-1; 138,0 106 s-1 ]
f [4,9 MHz;6,7 MHz] α [0,5 106 s-1; 5,97 106 s-1] β [0,9 106 s-1; 54,8 106 s-1]
-
-
-
-
Overlaped PD patterns
Cluster #1
β [2,1 106 s-1; 29,8 106 s-1]
Cluster #2
f [2,0 MHz; 7,2 MHz] α [0, 58 106 s-1; 86,5 106 s-1] β [0,4 106 s-1; 117,0 106 s-1 ]
Cluster #3
Page 90
f [18,4 MHz; 25,5 MHz] α [1,8 106 s-1; 160,0 106 s-1 ] β [3,1 106 s-1; 130,0 106 s-1]
Guidelines for PD Detection
The three sensors detected cluster #1 which results from an internal insulation defect within the transformer. The highest sensitivity was achieved with the sensors connected to the capacitive taps. The three sensors also detected cluster #2 which is related to corona and surface PDs. Only the HFCT sensor, connected to the tank earth (PD_S3), was able to detect the pulses related to the disturbances generated by external power electronics. After opening of the transformer, traces of damage due to insulation overheating was observed (see Figure 106).
FIGURE 106: INTERNAL DEFECT DETECTED. INSULATION DAMAGE DUE TO HEATING
4.5.2 Case Study 2: Combination of conventional and UHF PD detection methods The transformer investigated was a 5/7 MVA ONAN/ONAF 66000/22000 V Dyn1 unit constructed in 1998 (Figure 107). It was initially put into service in an urban environment where it operated predominantly on a fixed tap and without any problems until removed from service in 2007 as part of a major upgrade project. Later in 2007, the transformer was relocated to a rural environment where it was used to replace a failed transformer. The unit was not heavily loaded and gave no indication of any problems until February 2010, when the result of the annual dissolved gas analysis (DGA) run on an oil sample indicated 5400 ppm of hydrogen. The dissolved gas ratios indicated that PD was taking place [62]. In order to better monitor what was occurring, an on-line DGA monitor was installed. Initially the monitor recorded a lower concentration of H2, but it was quickly realized that this was actually a false reading due to the fact that the detection limit for a dissolved hydrogen level of 3000 ppm had been exceeded and, counterintuitively, this was causing the monitor too read low. The transformer was then partially de-gassed. The results for the next three years are shown in Figure 108. As can be seen, the hydrogen and methane levels showed a steady increase, which seemed to indicate that PD activity was ongoing.
Page 91
Guidelines for PD Detection
FIGURE 107: 15/7 MVA, 66/22 kV TRANSFORMER
FIGURE 108: LONG-TERM ON-LINE DGA RESULTS OF 15/7 MVA, 66/22 kV TRANSFORMER
Two methods were recommended for PD measurements: the conventional method (according to IEC 60270) with sensors installed at the bushing taps (Figure 109), and an unconventional ultra-high frequency (UHF) method with a sensor placed inside the transformer tank (Figure 110). With the conventional method, the PD signal from each tap was synchronously acquired by a three channel acquisition unit. The center frequency of the digital band pass filter of the acquisition unit being selected to reach the best signal-to-noise ratio. To obtain more detailed information about the type and location of the insulation PD defects, unconventional UHF PD measurements in the frequency range between 100 MHz and 2 GHz with an antenna installed inside the oil drain valve were performed. The presence of external noise in this frequency range is low, and external radio or mobile phone signals are easily recognized and eliminated from the measurements. PD activity inside the bushing insulation and close to the end winding area is mostly detected with the conventional method, while the rest of the tank is covered by the UHF antenna. Even if the PD signal is measured in a different frequency range, the PRPD patterns obtained exhibit strong similarity and thus make recognition of the defect type easier. The signal detected by the UHF antenna was synchronized with the signal detected at the bushing taps. Furthermore, the pulses measured in the UHF range, mostly coming from internal PD activity, can trigger the
Page 92
Guidelines for PD Detection
start of conventional measurements. Thus a better separation between internal and external PD pulses can be obtained.
FIGURE 109: TAP SENSOR FOR PD MEASUREMENTS
FIGURE 110: UHF ANTENNA INSTALLED IN OIL DRAIN VALVE
The three-phase PD trend is presented in Figure 111. At each point indicated in the trend diagram, PRPD patterns and 3PARD [63] are available. Figure 112 depicts the PRPD patterns of the PD signal acquired by the three-channel synchronous system. They are complex patterns with several overlapping PD sources. In order to separate clusters of different PD sources, a synchronous multi-channel PD evaluation technique was applied. The 3PARD diagram (Figure 113) visualizes the relationship between amplitudes of a single PD pulse in one phase and its crosstalk-generated signals in the other two phases. By repetition of this procedure for a large number of PD pulses, PD sources within the test object as well as external noise sources appear as clearly distinguishable concentrations or ‘clouds’ of pixels in the 3PARD diagram. By examining individual clusters in the 3PARD diagram, a separation between noise and PD phenomena is possible [63], [64]. Figure 114 shows the transformation back to PRPD pattern from clusters 1 and 2. The patterns of clusters 1 and 3 appear to be generated by bubbles and surface discharge, with the highest amplitude in phase B (cluster 1) and phase A (cluster 3). Similar PRPD patterns were reported in [65]. The shape and phase position of the patterns of the clusters 2 (phase A) and 6 (phase C) may indicate partial discharge activity inside voids within the insulation system. The clusters 4 and 5 are generated by external interference.
FIGURE 111: THREE PHASE PD TREND OF 15/7 MVA, 66/22 kV TRANSFORMER
Page 93
Guidelines for PD Detection
FIGURE 112: THREE PHASE SYNCHRONOUS PRPD PATTERNS FROM 15/7 MVA, 66/22 kV TRANSFORMER
Two spectra of the signal were obtained by performing a frequency sweep (Figure 115). The upper spectrum is built up based on the maximum peak amplitudes of the time domain signal acquired at each value of the frequency during the sweep. The lower spectrum corresponds to their minimum amplitudes. Internal PD activity is always visible on the upper spectrum while external interferences (corona discharge, radio waves, GSM) are visible on both spectra.
Phase C
FIGURE 113: EQUIVALENT 3PARD DIAGRAM
Phase B
Phase A
FIGURE 114: INDIVIDUAL PRPD PATTERNS OF THE SELECTED CLUSTERS
Internal PD activity was identified in the frequency range from 450 MHz to 650 MHz. The PRPD pattern corresponding to a center frequency of 600 MHz is presented in Figure 116. The signal was synchronized with a 50 Hz signal taken from the measuring tap of phase A. It can be seen in Figure 114 that the phase of the voltage where the PDs occur is the one which is characteristic to internal discharges. Furthermore, it indicates a possible location of the PD activity, namely, in the vicinity of phase A. The PRPD patterns of frequencies between 1 GHz and 1.4 GHz were checked and no internal PD activity was found.
FIGURE 115: FREQUENCY SWEEP DIAGRAM
FIGURE 116: PRPD PATTERN
Page 94
Guidelines for PD Detection
4.5.3
Case Study 3: Combination of conventional and UHF partial discharge detection methods The transformer in question was a 130/130/100 MVA – 230/115/48 kV unit constructed in 1973 (Figure 117). The PD monitoring system was installed in 2013 on the 230 kV RBP bushings. For each bushing, one capacitive tap PD sensor with multiple redundant protection was installed (Figure 118) and connected synchronously to an acquisition unit. A UHF antenna was installed within the upper part of the transformer tank (Figure 119) and was first connected to the UHF down-converter and later to the acquisition unit.
FIGURE 117: 130/130/100 MVA – 230/115/48 kV TRANSFORMER
FIGURE 118: PD TAP SENSOR
FIGURE 119: UHF SENSOR PLACED BETWEEN PHASES V AND W
The three-phase PD trend is presented in Figure 120. For each point of the trend, PRPD patterns and 3PARD diagrams are available. Figure 121 depicts the PRPD patterns of the PD signal acquired in April 2014. The patterns are complex, with several PD sources overlapping each other. In order to separate clusters of different PD sources, a synchronous multi-channel PD evaluation technique (3PARD) is applied [63]. The back transformation to PRPD patterns of the clusters 1 and 2 is presented in Figure 121. The pattern of cluster 1 indicates the presence of PD activity inside gas cavities. The highest amplitude of the signal is detected at the phase V, but signal cross-talk to phases U and W is also visible. The PRPD pattern of cluster 2 appears to be generated by PD surface discharge in the vicinity of the phase W. The other clusters visible in the 3PARD diagram are generated by external interference.
Page 95
Guidelines for PD Detection
3PARD diagram
3-phase PRPD patterns
FIGURE 120 THREE PHASE PD TREND FROM THE 130/130/100 MVA – 230/115/48 kV TRANSFORMER
FIGURE 121: SEPARATION OF PD SOURCES USING 3PARD FROM THE 130/130/100 MVA – 230/115/48 kV TRANSFORMER
In order to gain information about the frequency content of the acquired PD signal, a frequency sweep is performed. Both off-line and on-line sweeps are presented in Figure 122 and Figure 123, respectively. An offline frequency sweep was performed during the installation of the monitoring system, while the transformer was de-energized. Such an off-line spectrum provides important information about the sources of interference produced by other equipment in the substation and vicinity. These sources must be discarded when the analysis of the on-line detected PD signal is performed. The increasing upward trend line of the UHF PD signal and PRPD pattern of the signal at 560 MHz are shown in Figure 124. An on-line frequency sweep was performed and internal PD activity was identified in the frequency range from 450 to 650 MHz (Figure 123). The signal detected during the on-line monitoring was synchronized with a voltage signal taken in phase U. The PRPD patterns at frequencies above 1 GHz were also investigated and no PD activity was identified.
Page 96
Guidelines for PD Detection
FIGURE 122: ON-LINE FREQUENCY SWEEP DIAGRAM
FIGURE 123: ON-LINE FREQUENCY SWEEP DIAGRAM
FIGURE 124: TREND DIAGRAM AND PRPD PATTERNS OF THE SIGNAL DETECTED IN THE UHF RANGE
Three independent sources of the received PD signal are necessary to build a 3PARD diagram. In Figure 125, such a diagram was built up using the UHF signal together with the two conventional signals measured at the bushings of phase V and W. After the cluster separation, the PD activity was confirmed and the PRPD patterns were identified from the clusters surrounded by the red rectangles shown in Figure 125. They have similar shape and voltage phase position to those presented in Figure 121. Detecting the PD activity in the UHF range is a good indication that the PD source is located inside the transformer tank and not inside the bushing.
Page 97
Guidelines for PD Detection
FIGURE 125: PD SOURCES SEPARATION COMBINING THE CONVENTIONAL AND UNCONVENTIONAL MEASUREMENTS
PD activity was detected at phases V and W of the transformer by conventional and UHF measurements. The evolution of the PD activity is presented in Figure 126. It can be noticed that the amplitude of the PD pulses increased by a factor of 3 over time. The operating conditions of the transformer were similar and did not influence the readings of the PD pulses amplitude. Dissolved Gas Analysis (DGA) was also performed (Table 8). The increase of the H2 and CH4 concentrations confirms the presence of the PD activity. The increase of the CO concentration indicates paper deterioration, probably as an effect of the PD. TABLE 8: FAULT GASES CONCENTRATIONS FROM LAB TESTS Sample date
H2
CO
CO2
CH4
C2H2
C2H4
C2H6
N2
O2
15-04-2014
433
416
3016
115
9.1
92
15.4
33680
1100
15-05-2014
966
835
5952
226
21.4
179
31.5
60120
860
12-06-2014
1212
808
5797
225
20.5
171
30.4
65440
1390
Phase V
Phase W
FIGURE 126: INCREASE OF THE AMPLITUDE OF THE PD PULSES DETECTED AT THE PHASES V AND W
In order to obtain a more precise localization of the PD source, acoustic PD measurements were performed on the1st of July. Figure 127 shows a preliminary position of the sensors. After the repositioning of acoustic sensors (S1…S4), the coordinates of the PD source were determined (Figure 128 – red dot). According to the results of the acoustic method, the PD activity takes place at the exit leads of the HV winding of phase V.
Page 98
Guidelines for PD Detection
For the next step, the transformer was de-energized and the bushing of phase V was dismantled (Figure 129). Internal inspection was performed using an endoscope (Figure 130) and the PD activity around the phases V and W was confirmed – Figure 131 and Figure 132 respectively.
FIGURE 127: ACOUSTIC PD LOCALIZATION
FIGURE 129: DISMANTLED BUSHING OF PHASE V
FIGURE 128: PD SOURCE LOCALIZATION IN THE VICINITY OF THE PHASE V
FIGURE 130: ENDOSCOPIC INSPECTION
Page 99
Guidelines for PD Detection
FIGURE 131: PD TRACKS AT PHASE V
FIGURE 132: PD TRACKS AT PHASE W
4.5.4 Case Study 4: Combination of conventional method with acoustic method A 16 MVA, 150/20.8 kV power transformer of dimensions 4.25 m x 1.36 m x 2.15 m was subjected to dielectric routine tests (Figure 133) [66]. The PD level has been measured using the conventional method by connecting a synchronous three channel PD acquisition unit to the capacitive taps of the transformer HV bushings (Figure 133). After calibration, the measurements were performed at a center frequency of 500 kHz with a 650 kHz bandwidth. The testing voltage has been linearly increased, and at 85 kV, a PD cluster characterized by about 4 nC of apparent charge has been detected in phase W (Figure 134). The same type of pattern but of lower apparent charge magnitude was measured in the other two phases. Due to the high PD level on phase W, it is assumed that the PD is located near or inside the winding of phase W and the PD activity on phase U and V is only measured due to inductive and capacitive coupling between the phases. For more accurate localization of the PD source, acoustic measurements were undertaken. For the acoustic measurements, four acoustic emission sensors with resonant frequency of 150 kHz were placed in different location on the exterior of the transformer tank. Using the difference in arrival time of the acoustic PD signal at multiple sensors, the appropriate software computes the location of the PD source [66], [59]. These time differences are the only available data from the acoustic method, and their accuracy determines the precision with which the PD defect(s) can be located. Higher accuracy of a defect’s localization can be obtained by triggering the acoustic signals with conventional PD signal. This means that the acquisition of the acoustic signals is started only if the electrical PD signal is detected. Through stepwise repetition of the measurement, the acoustic sensors were placed progressively closer to the assumed location of the defect. The signal acquired by each sensor is presented in Figure 135. Arrival times and voltage amplitudes indicate a defect location in the upper part of the low voltage side of the winding of phase W. The calculation of the defect location resulted in the coordinates x = 2.6 m, y = 1.1 m and z = 1.4 m. This location is related to the position of a shield electrode in the low voltage side of the winding of phase W. All three low voltage phases were energized, one after the other, and acoustic PD was only detected when phase W was energized. After opening the transformer tank, a foreign copper wire was found at the indicated location.
Page 100
Guidelines for PD Detection
FIGURE 133: MVA 150/20.8 kV POWER TRANSFORMER AND PD DECOUPLING AT THE BUSHING TAP
FIGURE 134: CONVENTIONAL PD MEASUREMENTS AT U= 85 kV IN PHASE V (UPPER PATTERN) AND IN PHASE W (LOWER PATTERN)
Page 101
Guidelines for PD Detection
FIGURE 135: ACOUSTIC MEASUREMENTS AND DEFECTS LOCATION
The drawing on the left shows the circular intersections which result from the automatically computed arrival times for all three sensors; these circles indicate the probably location of the PD source.
4.5.5 Case Study 5: Localization of PD by acoustic and UHF-measurement Because of increasing gas-in-oil values, a 333 MVA, 400/220 kV single-phase autotransformer was tested onsite and on-line for PD [67]. The high noise level at site strongly disturbed the conventional PD measurements made according to IEC 60270 at frequencies lower than 1 MHz. In this case the main source of the noise was the 400 kV bus bar above the transformer, which was producing audible corona discharges. Consequently, UHF PD measurements for PD detection in combination with acoustic measurements for PD location were performed in order to get reliable results. In this case, the transformer contained three oil filling valves and thus three identical UHF Sensors were able to be installed, one in each valve. Figure 136 shows the positions of the UHF sensors (UHF 1 – UHF 3). Two sensors are opposite to each other at the top of both front ends of the tank and the third (UHF 3) is located at the bottom in the middle of the transformer side.
Page 102
Guidelines for PD Detection
FIGURE 136: 333 MVA TRANSFORMER SHOWING POSITIONS OF UHF SENSORS AND ACOUSTIC SENSORS [67]
First, the so-called dual port performance check was done [68]. Artificial UHF impulses were injected at each sensor with a signal generator (60 V at 50 Ω). It was not possible to detect the artificial impulses at any combination of emitting and receiving sensor. The manufacturer’s design drafts explain this strong damping of the UHF signals: tubes are installed behind the oil filling valves in order to direct the oil flow around the winding. According to the unsuccessful dual port performance check, it could be stated that the sensors are electromagnetically decoupled from each other and might also be shielded against UHF pulses from internal PDs. A further explanation might be that the maximum signal generator output voltage of 60 V is not sufficient to transmit UHF waves through that particular transformer. Nevertheless, UHF signals from internal sources were detectable at all three sensors at nominal voltage, i.e. the internal PD was producing UHF signals with higher energy content than the applied artificial impulses. It can be concluded that the Dual Port performance check is thus just a worst-case estimation of the sensitivity. But even though the performance check was not successful, sensitive UHF measurements were still possible. Frequency analysis of the signals measured from the installed UHF probes revealed the internal shielding characteristic of the tank, see Figure 137. The signals exhibit frequency content of up to 1 GHz, as emitted by a broadband emitter of UHF waves, similar to and typical of internal PD in oil. External disturbing sources would have been narrow band, e.g. at around 500 MHz for digital video broadcasting or around 900 MHz or 1800 MHz for GSM I and II, respectively, all of which are modulated carrier frequencies. In Figure 137 the unamplified measured signals of the UHF probes are shown with their frequency analyses (FFT).
Page 103
Guidelines for PD Detection
FIGURE 137: MEASURED PROPAGATION TIME DIFFERENCES BETWEEN THREE UHF PROBES
Shown are the time-domain signals from the 3 UHF probes (unamplified) and their associated frequency-domain spectra. The latter exhibit distinctly broadband characteristics, indicating PD, without external EMI.
Propagation time differences in the range of nanoseconds (ns) are recognizable between the signals. Taking the propagation time differences caused by different lengths of measuring cables into account, a first estimation of the geometric PD location pointed to the on-load tap changer (OLTC) on the left hand side of the transformer. That is supported by the measured UHF amplitudes of the three UHF probes. The probe nearest to the tap changer (probe UHF 2) show the highest output reading of 10 mV, whereas the other probes did not reach more than 5 mV. Therefore probe UHF 2 was used for triggering and determining the starting time in order to calculate the propagation time differences. PD also produces acoustic waves, which are measured with piezo-electric sensors installed on the outer tank wall. Their measurement frequency range is between 50 and 200 kHz. Due to comparatively high acoustic signal attenuation within the solid and liquid insulation material along with intervening structures inside the transformer, sensitive acoustic measurements are hard to achieve [57]. Additionally, acoustic signals of PD might be obscured by higher amplitude signals produced by ambient mechanical noise and inherent noises within the transformer itself (core noise). Summarizing, exclusive acoustic PD measurement is only useful to a limited extent. To increase the sensitivity of acoustic measurements the method is combined with the more sensitive UHF measuring method. UHF signals are used as trigger signals in order to activate the acoustic measurement during the occurrence of UHF PD signals. By using averaged signals (averaging in time domain), the acoustic PD pulses add up constructively whereas the white background noise is averaged to zero. Thus the signal-to-noise ratio (SNR) of the acoustic signals is increased by implementing this technique. The UHF measuring method is based on picking up electromagnetic waves which radiate in all directions from the PD source, at approximately two-thirds of speed of light, inside the transformer. Therefore, for the purpose of acoustic location, the UHF signals are detected at essentially the same time the PDs occur,
Page 104
Guidelines for PD Detection
enabling the use of the UHF signals as a trigger for the acoustic measurements. Conversely, the speed of acoustic waves is 1400 m/s [54], producing transit times within the range of milliseconds. Geometrical distances between sensors and the source of PD (calculated from the propagation times of the individual acoustic sensors) result in a spherical region inside the transformer. With at least three acoustic sensors and the corresponding propagation times, it is possible to calculate the intersection of the spheres and thus to determine the PD location. It must be assumed that the acoustic waves travel directly along the line-of-sight path from the PD source through the oil and through the steel tank to the sensor without any reflections. However, the location process has also to deal with acoustic waves travelling faster through the tank wall (whose propagation velocity is much higher) than through the oil. The propagation times of the acoustic signals can be computed objectively with the help of the Hinkley criterion [67]. It is based on the signal energy of the measured signal and results in an absolute minimum for the signal starting point. Figure 136 depicts the positions of the acoustic sensors used (A1 – A6). As an example, Figure 137 shows the measured and averaged acoustic signals of the acoustic sensors A5 and A6. Averaging was performed with approx. 100 signals. Although averaging was used, the SNR was quite low, thus the determination of the propagation times was only possible by application of the Hinkley criteria. The respective propagation times are tA5 = 1.03 ms and tA6 = 2.07 ms. Based on the determined propagation times, the supposed position of the PD source is located in the vicinity of the tap changer (Figure 138) [68]. Geometrical inaccuracy is within the range of approx. 40 cm on all space axes. This inaccuracy is caused by using different combinations of propagation time differences and different location methods [59].
FIGURE 138: DETERIORATED PAPER INSULATION ON LEADS AT THE TAP CHANGER
After transportation of the transformer to a factory, the location result was confirmed by an IEC triggered acoustic measurement in the test area, so the transformer was detanked for inspection and repair. The visual inspection of the tap leads at the tap changer region confirmed the location results and revealed deteriorated paper insulation, see Figure 138. After repair of the affected leads, the transformer passed the acceptance test without any indication of PD activity and was put back into service. 4.5.6 Case Study 6: Monitoring by UHF PD-measurement Online monitoring of power transformers, which supports established diagnosis methods, is steadily gaining acceptance. Continuous measurement using trend analysis allows the detection and tracing of undesirable changes at an early state. For PD, ultra-high frequency monitoring represents an advantageous technique, because the measurement is done inside the tank and is thus much less sensitive to external noise. Additionally, it is applicable to transformers in service. The considerable amount of data generated requires appropriate evaluation; partially automated analysis is inevitable. A 50-year-old unit generator transformer with a rated voltage of 110/10 kV and a rated power of 120 MVA was monitored [69]. An online UHF PD measurement system recorded data. UHF PD signals were measured with approximately 35 dB amplification and a bandwidth of 9 MHz at a center frequency of 505 MHz. The noise level of the system was between 1.5 and 2 mV. Therefore, all PD below 2 mV were discarded. Phase L1 was used for phase correlation. Due to the fact the generating unit is only in operation on demand, the transformer is not continuously in service. Measurements
Page 105
Guidelines for PD Detection
are available for approximately 65 days from 2009 until 2012. This case study presents an approach using phase-resolved PD pattern analysis. Typical patterns from known PD sources are reduced to an abstracted shape which uniquely characterizes the form of the PD source. This so called template is compared to measured PRPD patterns gained from the monitoring data. Comparison between pattern and template is calculated by 2dimensional normalized cross-correlation algorithm. Source tracking over time is evaluated using continuous correlation. By introducing a set of templates for correlation, the progress of individual PD sources is determined. Cross-correlation is an algorithm used for pattern recognition within an image. The higher the similarity between two images, the higher is their correlation factor. In this contribution, the normalized cross-correlation is used providing values of correlation coefficients between -1 and +1 for each matrix element. Thus, cross-correlations of different images become comparable. A correlation coefficient of 1 indicates an exact match of the template (this never occurs in practical pattern analysis). -1 represents an area where image intersection and template are opposed (negative image, also never occurs in practical pattern analysis). Figure 139 shows three typical PD patterns of this transformer.
FIGURE 139: UHF PRPD PATTERNS 1 – 3 [69]
Patterns should be traceable over time. Therefore, the constant PD data stream is divided into segments with constant duration. For each segment, the PRPD pattern is generated and then cross-correlated with a template. Determining an adequate duration period depends on the behavior of the source over time. Pattern 1 from Figure 139 shows high volatility. Therefore, duration is set to 1 minute. Correlation is calculated with the small template from Figure 139. The maximum value of the correlation matrix represents the correlation coefficient for the time segment. An example is plotted in Figure 140. The correlation coefficient is shown in red, aligned to the left axis. For comparison, the number of PDs per minute is also plotted (black, right axis).
Page 106
Guidelines for PD Detection
FIGURE 140: NUMBER OF PD PER MINUTE AND CORRELATION COEFFICIENT [69]
Figure 141 shows a graph of the frequency of occurrence for each pattern. The colors indicate each time the pattern types shown in Figure 139 were detected by cross-correlation. Pattern 1 is only present at 15% of the time showing intermitted behavior. Pattern 2 has a higher rate of appearance and can be detected at 40% of the measurement time. Pattern 3 is the dominating source which can be detected 60% of the time over the entire period.
FIGURE 141: RESULTS OF RECOGNITION ALGORITHM OF DETERMINED PATTERNS 1 - 3 [69]
PD data can be evaluated using normalized cross-correlation for PRPD pattern recognition. Therefore, the characteristic shape of a pattern is defined by a template matrix. The data is segmented into constant periods of time which are used to generate PRPD patterns. Each pattern is cross-correlated with the template. The result of a correlation is a matrix whose coefficients define the similarity between template and pattern. The quality of correlation depends on several parameters. Filtering can improve correlation outcome; results strongly depend on the chosen filter method and its parameterization. If preconditions are met, a threshold level can be defined. The maximum value of the resulting matrix is used as trigger indicating the presence of the determined pattern. Its x-coordinate represents the phase angle the pattern occurs. Combination of both allows long term tracking of PD and thus its behavior over time. Comparison of pattern shape and phase angle with known PD sources from literature can be made. In this contribution, only the relative phase can be considered due to the UHF measurement. Nevertheless, the cross-correlation method presented here can be applied to any PD measurement method which arrays the PD signal pulses against the phase angle of the applied voltage. The essential benefit of the method presented is its application on large PD datasets, e.g. from monitoring systems. In the case presented, 65 days of monitoring data was evaluated. Therefore, three PRPD patterns Page 107
Guidelines for PD Detection
being typical for the determined transformer are evaluated. Using cross-correlation, it is possible to track patterns over the monitoring period in terms of their appearance and their phase positions.
4.6
Advantages and disadvantages of different methods
In Table 9, UHF and conventional (IEC 60270) measurement methods are compared with respect to specifically defined and chosen topics. TABLE 9: COMPARISON OF UHF AND CONVENTIONAL IEC 60270 MEASUREMENT METHOD
IEC 60270
UHF
1) Actual PD source level
pC
not known
mV
Not known
2) Attenuation of coupling path
Low pass filtering of winding & ratio of internal capacitances
Not known
Damping of EM waves
Not known (but very low)
3) Sensor sensitivity
Ratio coupling capacitor specimen capacitance
Calibrated
Antenna factor
Not yet calibrated
Installed sensor
Sensitivity check in preparation by WG A2/D1.51
< 100 pC of apparent charge
@1,2xUm/√3
-
-
4) Acceptance test level
4.7
Summary and important aspects
1) It is emphasized that when making PD measurements using unconventional methods, it is not possible to quantitatively correlate or calibrate the received signal level in terms of PD charge(pC or nC). 2) As also shown through practical measurements on a winding model, conventional PD measurment according to IEC 60270 may also be affected owing to damping (attenuation) caused by the low pass behavior of the winding. Therefore the calibration routine may not eliminate all influences. It is beneficial if lower frequencies are used during the conventional PD measurement, which is also noted in IEC 60270 and IEC 60076-3. Nevertheless it is well known that conventional PD measurement is also often done at higher frequencies (e.g. to improve SNR); in such cases it is important that the low pass filtering effect of the windings should not be neglegted. 3) It is well known that PD events in insulation oil exhibit steep rise-times and thus their frequency content extends to very high frequencies. In actual research results (by using state-of-the-art extra high frequecy measuring devices) the rise time was measured down to several 10 ps. 4) Unconventional UHF PD measurement is beneficial because the transformer vessel acts as a Faraday shield for ultra high frequencies, so that external corona noise and other sources of external electromagnetic intereference (EMI) is filtered out, although some external signal can still enter via the bushings. UHF PD measurements on transformers can often result in high SNR values, whereby damping and sensor effects must always be considered. Page 108
Guidelines for PD Detection
5) Unconventional acoustic PD measurement is well known for localization of PD event using several sensors and measuring the time difference between the acoustical signals. However, the SNR may in different cases not be sufficiently high enough to be able to accurately estimate the exact arrival times of the incoming signals. By using averaging techniques together with a reference signal (UHF or conventional PD measurement signal) for a trigger reference, the SNR of the acoustical signals can be significantly improved and thus their arrival times can be more accurately estimated, leading in turn to more accurate localization of the discharge site. 6. Again, it is good to follow general best practise procedures when carrying out all PD measurements, i.e. extensively documenting test set-ups and taking detailed notes and so on. This is especially important when employing unconventional methods because the test set-ups are not industry standardized, but typically assembled according to the needs of the user and the requirements of the equipment at hand. In this chapter, an overview about PD-measurements on transformers employing both conventional and unconventional methods is given and several practical examples are presented showing the application of different measuring systems working on different measurement principles. The progress in PD measurements on transformers is currently under discussion in CIGRE WG D1.29. The attempt has been made to define possible criteria how to distinguish between dangerous and less dangerous PD-sources in oil-impregnated electrical insulation system of power transformers. Despite a large number of practical examples showing the identification and localization of PD-sources, unambiguous identification of PDsources in the electrical insulation system of transformers together with a reliable estimate of the actual risk they present remains a topic for further research.
Page 109
Guidelines for PD Detection
5 REFERENCES [1] L. Niemeyer, “A generalized approach to partial discharge modelling,” IEEE Trans. Diel Elec Insul., Vols. 2, pp. 510-28, 1995. [2] U. Schichler et al., Risk Assessment on Defects in GIS Based on PD Diagnostics (TB525), CIGRÉ AG D1.03, 2013. [3] CIGRE WG 15.03, “Effects of Particles on GIS and the Evaluation of Relevant Diagnostic Tools,” in CIGRE Session, Report 15-103, Paris, 1994. [4] H. Okubo, N. Hayakawa and A. Matsushita, “The Relationship Between Partial Discharge Current Pulse Waveforms and Physical Mechanisms,” IEEE Electrical Insulation Magazine, Vols. Vol. 18 No. 3 pp. 38-45, 2002. [5] E. Gulski et al., “Guidelines for Unconventional Partial Discharge Measurements (TB444),” CIGRÉ WG D1.33, Paris, 2010. [6] S. Boggs, “Electromagnetic Techniques for Fault and Partial Discharge Location in Gas-Insulated Cables and Substations,” IEEE Trans. on Power Apparatus and Systems, Vols. PAS-101, No. 7, pp. 1935-1941, 1982. [7] B. Hampton and R. Meats, “Diagnostic Measurements at UHF in Gas Insulated Substations,” IEE Proceedings C: Generation, Transmission and Distribution, Vols. 135, Issue 2, pp. 137 – 145, 1988. [8] J. Pearson, 0. Farish, B. Hampton, M. Judd, D. Templeton, B. Pryor and I. Welch, “Partial Discharge Diagnostics for Gas Insulated Substations,” IEEE Transactions on Dielectrics & Electrical Insulation, vol. 2 No. 5, 1995. [9] M. Judd, O. Farish and B. Hampton, “The Excitation of UHF Signals by Partial Discharges in GIS,” IEEE Transactions on Dielectrics, Vols. 3, pp.213-228, 1996. [10] CIGRE JWG 33/23.12, “Insulation Co-ordination of GIS: Return of Experience, On Site Tests and Diagnostic Technique,” Electra No.176, pp. 67-97, February 1998. [11] CIGRE WG 15.03, “Diagnostic Methods for Gas Insulating Systems,” in Report 15/23-01, CIGRE Session, Paris, 1992. [12] A. Reid and M. Judd, “Ultra-wide bandwidth measurement of partial discharge currents in SF6,” Applied Physics, 2012. [13] IEC60270, “High-voltage test techniques – partial discharge measurements,” 2000. [14] R. Pietsch, W. Hauschild et al., “High-Voltage On-Site Testing with Partial Discharge Measurements (TB502),” CIGRÉ WG D1.33, Paris, 2012. [15] A. Bargigia, W. Koltunowicz and A. Pigini, “Detection of Partial Discharges in Gas Insulated Substations,” IEEE Transactions on Power Delivery, Vols. 7, No. 3, pp. 1239-1249, 1992.
Page 110
Guidelines for PD Detection
[16] M. Leijon and A. Vlastos, “Pattern Recognition of Free Metallic Particles Motion Modes in GIS,” in IEEE Inter. Symp. Elec. Insul., Boston, 1988. [17] L. Lundgaard, “Partial Discharge - Part XIII: Acoustic Partial Discharge Detection - Fundamental Considerations,” IEEE Elec Insul Mag, Vols. Vol.8, No.4, 1992. [18] L. Lundgaard, “Partial Discharge - Part XIV: Practical Application,” IEEE Elec Insul Mag., Vols. 8, No.5, 1992. [19] L. Lundgaard, “Particles in GIS Characterization from Acoustic Signatures,” IEEE Transactions on Dielectrics & Electrical Insulation, vol. 8 No.6, 2001. [20] IEC62478, High-voltage test techniques – measurement of partial discharge by electromagnetic and acoustic methods (publication pending). [21] CIGRÉ TF 15/33.03.05, “Sensitivity Verification for the UHF and Acoustic Method,” Electra, No. 183, pp. 75-87, April 1999. [22] S. Hoek and S. Neuhold, "Tuned Medium-Band UHF PD Measurement Method for GIS," in Cigre Session, Paris, France, 2012. [23] Agilent, “Spectrum Analyzer Basics, Application Note 150,” Agilent Technologies Inc.,5952-0292, CA, USA, 2004-2006. [24] M. Hikita, S. Ohtsuka, T. Hoshino, S. Maruyama, G. Ueta and S. Okabe, “Propagation properties of PDinduced electromagnetic wave in GIS model tank with T branch structure,” IEEE Trans Dielectrics & Electrical Insulation, Vols. 18, no 1, pp 256 – 263, 2011. [25] M. Hikita, S. Ohtsuka, J. Wada, S. Okabe, T. Hoshino and S. Maruyama, “Study of partial discharge radiated electromagnetic wave propagation characteristics in actual 154 kV model GIS,” IEEE Transactions Dielectrics & Electrical Insulation, Vols. 19, No. 1, pp. 8-17, 2012. [26] S. Neuhold, T. Heizmann, R. Bräunlich, D. Köchli, U. Riechert and C. Dehne, “Experience with UHF PD detection in GIS using external capacitive sensors in windows and disk-insulators,” in XVth International Symposium on High Voltage Engineering (ISH), Ljubliana, 2007. [27] S. Hoek, U. Riechert, T. Strehl, S. Tenbohlen and K. Feser, “A New Procedure for Partial Discharge Localization in Gas-Insulated Switchgear in Frequency Domain,” in 14th ISH, Beijing, China, 2005. [28] R. Pietsch, “Sinnvoller Einsatz von UHF-Teilentladungsmessungen bei Entwicklungs- und VorOrt-Prüfungen,” in Haefely Symposium, Stuttgart, 2000. [29] F. Wester, Condition assessment of power cables using partial discharge diagnosis at damped AC voltages, Dissertation TU Delft, Optima Grafische Communicatie, ISBN:90-8559-019-1, 2004. [30] CIGRE WG 21.09, “After Laying Tests,” Electra, no. 166, 1996. [31] M. Wild, S. Tenbohlen, E. Gulski and R. Jongen, “Power cable modeling for PD pulse propagation and sensitivity,” in IEEE Electrical Insulation Conference (EIC), 2013. [32] M. Wild, S. Tenbohlen, E. Gulski, R. Jongen and F. d. Vries, “Practical aspects of PD localization for long
Page 111
Guidelines for PD Detection
length power cables,” in IEEE Electrical Insulation Conference (EIC), 2013. [33] F. Garnacho, M. Sánchez-Urán, J. Ortego, F. Álvarez and A. González, “Control of insulation condition of smart grids by means of continuous PD monitoring,” in CIRED, Paper 1103, Stockholm, Sweden, 2013. [34] Chan, Duffy, Hiivala and Wasik, “PD Testing of Solid Dielectric Cable,” IEEE Electrical Insulation Magazine, Vols. 7, No.5, 1991. [35] R. Plath et al., “PD Measurements on Extra High Voltage Cable Accessories During Commissioning Exeptionally Long 400 kV XLPE Insulated Cable System,” in International Symposium on High Voltage Engineering (ISH), 2007. [36] F. Garnacho, M. Sánchez-Urán, J. Ortego, F. Álvarez, D. Prieto, J. Vallejo and M. Jiménez, “Experiences of PD Measurements on HV Cable Systems Installed,” in CIGRE Session, Paper B1-206, Paris, 2014. [37] J. Fuhr, “Procedure for Identification and Localization of Dangerous PD Sources in Power Transformers,” IEEE Trans. on Dielectrics and Electrical Insulation, pp. 1005-1014, October 2005. [38] S. Ohtsuka et al., “Measurement of PD Current Waveforms in SF6 Gas with a Super High Frequency Wide Band Measuerement System,” in 16th International Symposium on High Voltage Engineering (ISH), 2011. [39] G. Cleary and M. Judd, “UHF and current pulse measurements of partial discharge activity in mineral oil,” IEE Proc.-Sci. Meas. Technol., vol. 153, no. No.2, 2006. [40] S. Ohtsuka et al., “PD Current Pulse Waveforms of Environmental Friendly Gases Measured with SHF_PDPW System and the Applicability of the UHF Method,” in IEEE International Conference on Condition Monitoring and Diagnosis (CMD), Indonesia, Bali, 2012. [41] M. Fukuzaki, S. Ohtsuka et al., “Frequency Bandwidth Dependence of PD current Waveforms in Transformer Oil Measured with the SHF_PDPW system,” in IEEE International Conference on Condition Monitoring and Diagnosis (CMD), paper G-22, Indonesia, Bali, 2012. [42] M. Fukuzaki, S. Ohtsuka et al., "Sophisticated Measurement of PD Current Pulse in Insulation Oil and the Effects of Degassing Treatment on the Waveforms with the SHF_PDPW System," in 18th International Symposium on High Voltage Engineering (ISH), PD-44, 2013. [43] S. Okabe, G. Ueta and H. Wada, “Partial discharge signal propagation characteristics inside the winding of gas-filled power transformer - study using the equivalent circuit of the winding model,” IEEE Transactions on Dielectrics and Electrical Insulation, pp. 1668-1677, 2011. [44] M. Siegel and S. Tenbohlen, “Comparison Between Electrical and UHF PD Measurement Concerning Calibration and Sensitivity for Power Transformers,” in CMD 2014, International Conference on Condition Monitoring and Diagnosis (CMD), Paper No. OB1-02, Jeju, Korea, 2014. [45] M. Judd, “Partial Discharge Monitoring for Power Transformers using UHF Sensors Part 2: Field Experience,” IEEE Electrical Insulation Magazine, vol. 21, no. 3, 2005. [46] S. Coenen, Measurement of Partial Discharges in Power Transformers Using Electromagnetic Signals, PhD Thesis, Universität Stuttgart, Germany, 2012. [47] A. Pfeffer, S. Tenbohlen and S. Kornhuber, “Influence of PD Location and Frequency Ranges on measured
Page 112
Guidelines for PD Detection
Apparent Charges,” in International Symposium on High Voltage Engineering (ISH), Hannover, Germany, 2011. [48] S. Okabe, G. Ueta, “Partial discharge criterion in AC test of oil-immersed transformer and gas-filled transformer in terms of harmful partial discharge level and signal transmission rate,” IEEE Transactions on Dielectrics and Electrical Insulation, pp. 1431-1469, 2012. [49] S. Coenen, M. Siegel, G. Luna, S. Tenbohlen, “Parameters influencing Partial Discharge Measurements and their Impact on Diagnosis, Monitoring and Acceptance Tests of Power Transformers,” in Cigre Colloquium, Paris, France, 2016. [50] S. Coenen, S. Tenbohlen, S. Markalous and T. Strehl, “Fundamental Characteristics of UHF PD Probes and the Radiation Behaviour in Power Transformers,” in ISH, Cape Town, South Africa, 2009. [51] M. Siegel, M. Beltle and S. Tenbohlen, “Characterization of UHF PD Sensor for Power Transformers using an Oil-filled GTEM Cell,” IEEE Transactions on Dielectrics and Electrical Insulation, p. (to be published), June 2016. [52] M. Siegel, S. Tenbohlen and S. Kornhuber, “Langzeitüberwachung von Leistungstransformatoren: Teilentladungsmonitoring von Leistungstransformatoren mit der UHF-Methode,” Elektrizitätswirtschaft (ew), Heft 25, Jg. 111 2012. [53] C57.127 IEEE Guide for the Detection and Location of Acoustic Emissions from Partial Discharges in OilImmersed Power Transformers and Reactors, New York, USA: The Institute of Electrical and Electronics Engineers, 2007. [54] E. Howells and E. Norton, “Parameters affecting the velocity of sound in oil,” IEEE Trans. Power App. Syst., Vol. 103, pp. 1111-1115, 1984. [55] M. Beyer, H. Borsi and M. Hartje, “Some aspects about possiblities and limitations of acoustic PD measurements in insulating fluids,” in 5th International Symposium on High Voltage Engineering (ISH), Braunschweig, Germany, 1987. [56] E. Großmann, Akustische Teilentladungsmessung zur Überwachung und Diagnose von Öl/Papier-isolierten Hochspannungsgeräten, PhD Thesis Universität Stuttgart, Germany, 2002. [57] S. M. Markalous, S. Tenbohlen and K. Feser, “Detection and Location of Partial Discharges in Power Transformers Using Acoustic and Electromagnetic Signals,” IEEE Trans. Dielectr. Electr. Insul., Vol. 15, pp. 1576-1583, 2008. [58] L. E. Lundgaard, W. Hansen and K. Dursun, “Location of discharges in power transformers using external acoustic sensors,” in Intl. Symposium on High Voltage Engineering (ISH), New Orleans, USA, 1989. [59] S. M. Markalous, Detection and Location of Partial Discharges in Power Transformers Using Acoustic and Electromagnetic Signals, Stuttgart: PhD Dissertation, Universität Stuttgart, 2006. [60] F. Garnacho and M. Sánchez-Urán, “PD Monitoring System for HV Substations,” in CIGRE Session, Paris, France, 2014. [61] F. Garnacho, M. Sánchez-Urán, J. Ortego, F. Álvarez, O. Perpiñan, E. Puelles, R. Moreno, D. Prieto and D. Ramos, “New Procedure to Determine Insulation Condition of High Voltage Equipment by Means of PD
Page 113
Guidelines for PD Detection
Measurements in Service,” in Cigre Session 2012, Paper D1-309, Paris, 2012. [62] L. Badicu, W. Koltunowicz, A. Piccolo, A. McGuigan and C. Feely, “Monitoring of a Distribution Transformer at Winchelsea Substation,” in Proceedings of TechCon 2014, Sydney, Australia, 2014. [63] W. Koltunowicz and R.Plath, “Synchronous Multi-Channel PD Measurements,” IEEE Transactions on Dielectrics and Electrical Insulation, Vols. Vol. 15, No. 6, p. 1715-1723, 2008. [64] K. Rethmeier, M. Krüger, A. Kraetge, R. Plath, W. Koltunowicz, A. Obralic and W. Kalkner, “Experience in on-site partial discharge measurements and prospects for PD monitoring,” in Proceedings of CMD 2008,, Beijing, China, 2008. [65] A. Carlson, J. Fuhr, G. Schemel and F. Wegscheider, Testing of Power Transformers – Routine tests, Type Tests and Special Tests, Zürich, Switzerland: 1st Edition, published by Pro Print, 2003. [66] S. Hoek, A. Kraetge, O. Kessler and U. Broniecki, “Time-based partial discharge localization in power transformers by combining acoustic and different electrical methods,” in International conference on Condition Monitoring and Diagnosis (CMD), Bali, Indonesia, 2012. [67] S. Coenen and S. Tenbohlen, "Location of PD Sources in Power Transformers by UHF and Acoustic Measurements," IEEE Transactions on Dielectrics and Electrical Insulation (TDEI), no. Vol. 19, Issue 6, pp. pp. 1934-1940, 2012. [68] S. Coenen, S. Tenbohlen, S. M. Markalous and T. Strehl, "Performance Check and Sensitivity Verification for UHF PD Measurements on Power Transformers," in Proceedings 15th Int’l. Symposium on High Voltage Engineering (ISH), Ljubljana, Slovenia, 2007. [69] A. Müller, M. Beltle, M. Siegel and S. Tenbohlen, “Assessment of UHF PD Monitoring Data by Means of Pattern Recognition,” in 18th International Symposium on High Voltage Engineering, ISH 2013, Paper OF405, Seoul, South-Korea, 2013.
Page 114