ECE 723 Report on Automatic Generation Control (AGC)
By Bingsen Wang
April 28, 2003
Implementation Issues of Automatic Generation Control Before and After Deregulation Bingsen Wang Department of Electrical and Computer Engineering University of Wisconsin – Madison 1415 Engineering Drive Madison, WI 53706, USA E-mail:
[email protected]
ABSTRACT
As an important aspect of the power system operation, the automatic generation control (AGC) has been extensively studied by my many researchers. In this paper the AGC is examined with focus on the practical issues related to the implementation. First, the basics of the AGC are reviewed and the implementation issues in the regulated environment are examined. Then the new issues to accommodate the dramatic utility structural changes due to deregulated market based operation are explored.
Key Words: Automatic Generation Control, Tie-line Bias Control, Load Frequency Control, Area Control Error, Deregulation
0. INTRODUCTION In North America, the interconnections consist of many control areas. In each control area, the generation must be controlled to meet the inside-area load (native load) in such a way that the system frequency can be maintained at the nominal one [1]. On the other had, due to the inter-area power transactions through the transmission lines between control areas (commonly called tie-lines), the generation in a control area must be regulated to take care of the scheduled tie-line transactions. Thanks to the changing nature of both native load and tie-line transactions, the generation level in each control area has to be changed from time to time. So the mechanical power input to some generators in a control area is regulated to keep balance between the total power generation and the power consumption. This regulation action is called automatic generation control. These generators are commonly referred to as units on regulation. In addition, the other functions of the AGC include: (1) regulate the system frequency deviation to zero; (2) maintain the net scheduled area interchange; (3) economically dispatch the generation among the generation sources. The economical dispatch can minimize the overall generation cost by rescheduling the entire system [2].
2
The schemes of AGC have evolved for years. And continued improvements of the schemes through application of new algorithms and technologies are expected [3]. In the last couple of decades, much effort has been focused on the AGC algorithms [4], [5], [6]. However, the implementation issues should be understood and paid enough attention for the AGC system to deliver good performance results. In sections I, II, III, and IV, the following implementation issues in the vertically integrated utilities’ settings will be presented in an order. •
Units’ response characteristics.
•
The selection of frequency bias in the tie-line bias control.
•
The filtering of the ACE signal to reduce the unnecessary control movements.
•
Post-installation/online tuning.
In section V, the possible adjustments for AGC to practically operate in the new deregulated environment are explored. Then the concluding remarks end this paper.
I. GENERATION UNITS’ RESPONSE CHARACTERISTICS Direct speed governing (primary control) and the supplemental adjustment of speed governor set points (secondary control) are the methods used on present day power systems for matching generation to load. If the system load increases suddenly, first the kinetic energy of the rotating parts in the system, such as rotors of generators, will be extracted to balance the load. Thus, the speed of the rotating parts across the system will decrease for a certain amount. This “natural response” can settle in one to two seconds. If the speed variation exceeds the deadband of the speed governor, which is usually 35 mHz, the speed governor will respond by increasing the turbine output. This control action is called primary control, which can usually complete in a dozen seconds. However the speed can not be pulled back to the one prior to the load increase. In order for the speed to return to the normal speed, the setpoint of the speed governor has to be changed. This follow up adjustment action is called secondary control, or automatic generation control. The simplified primary control and AGC are illustrated in Fig. 1. Typical speed droops for active governors are in the range of about 5% (which means 5% change in the speed will result in 100% change of the mechanic output power from the prime mover). Initially, the speed governor is operating on curve L 0 and the system is operating at the frequency ω0 (point A). After the load of the system frequency drops to ω1, the governor will respond by increasing the turbine output from P M0 to PM1 (point B). By changing the setpoint of the speed governor, the droop curve will move from L0 to L1. Now, the operation point is at C and system frequency has been brought back to ω0.
The AGC sends “raise/lower” signals/pulses to the regulated generation units to adjust the setpoint of the speed governor. So, the AGC performance will be definitely affected by the manner in which the generation units will respond to these control signals.
3
? (%)
5%
C
?0
A
?1
B
L1 L0
PM0
PM1
100%
PM (%)
Fig. 1 Speed governor droops characteristic
There are several types of generating units [3]: 1. Fossil-fired steam turbine Many drum type units are controlled in a boiler-following mode. The AGC pulses drive the speed-load setpoint adjuster on the speed governor, which, in turn, controls the steam flow at the turbine inlet. The boiler controls sense the changes of steam pressure and adjust the fuel-firing rate. The turbine under this mode can respond to AGC signals at the order of 3% per minute for a 30% excursion. For the once-through units, AGC signal is processed by a master controller that controls fuel, air, temperature, and turbine valve controls to limit undesired stresses on the plant components. A well adjusted un it of this type may be capable of making a 20% excursion in 10 minutes. 2. Nuclear steam units Nearly all nuclear plants have either boiling water (BWR) or pressurized water (PWR) steam generators. BWR units operated under AGC typically can respond at 3% per minute for 10 minutes within their control range. PWR units can make 20% excursion at rate of 3% per minute.
4
3. Combustion turbines and combined-cycle units Gas turbines are capable of fastest response of any units on utility system, but they are seldom operated under AGC since most of them are p eaking units. It’s more likely for combined-cycle units to be equipped with AGC control if they are operated by utility. Combined-cycle units operated by independent power producers are seldom operated under AGC. 4. Hydro units Lower-head The response characteristics of generation units will limit the rate of the AGC pulses. Even though the modern technology will give the AGC good capabilities of sending quite frequent control signals, it is not favored from either the feasibility of response or the considerations of operation cost. In fact, the ACE signal will be processed in such a way that the undesired control movements can be reduced, which will be discussed in section III.
II. SELECTION OF THE FREQUENCY BIAS COEFFICIENT IN THE TIE-LINE BIAS CONTROL The tie-line bias control, which was developed in mid thirties, is currently the AGC strategy of the interconnections in North America. The tie-line bias control is measured by the area control error (AEC), which is comprised of area imbalance between demand and supply (generation plus interchange), formulated as follows. ACE = ( Ni A – NiS ) – 10 β ( f A – f S )
(1)
Ni A: actual instantaneous net interchange (MW) , the algebraic sum of the power flows on the control area’s tie lines. Positive net interchange is a net power flow out of the control area. NiS : Scheduled net interchange (MW) , the mutually prearranged intended net power flow on the control area’s tie lines. Positive net interchange is a net power flow out of the control area. f A: actual system frequency (Hz), the actual frequency of the interconnection. f S: scheduled system frequency, the scheduled frequency in the interconnection, which is normally 60 Hz but may be offset to effect manual time error corrections. β : frequency bias setting (in MW/0.1Hz), the bias value used by the control area. 10: constant factor to convert frequency bias setting to MW/Hz.
Since the purposes of the automatic control are to keep the load balance in a control area and to keep the frequency at the normal frequency in an interconnection, there are two terms in the ACE: the first term is the generation and load mismatch; the second term is a function of the frequency deviation, known as area frequency response, which depends on the combined effects of droop characteristics of the speed governors and the frequency response of all loads. The area frequency response is the combination effects of the frequency dependency of both the generation and the load. The 10 B (F a – F s ) term represents the area contribution to the frequency regulation in the whole interconnection. If the frequency bias setting is set in such a way that the area frequency response exactly cancels the load mismatch term, the internal (to the
5
control area) disturbance can be distinguished from the external disturbance. For example, for the interconnected control areas as shown in Fig.2, all the interchanges are on schedule and the system frequency prior to the disturbance. Or in each control area, Ni A-NiS = 0 and F A – F S = 0. Suddenly, the load in control area “C” increases by sufficient amount, which results in the decrease of the system frequency, F A – F S < 0, across all control areas. In control areas “A”, “B”, and “D”, the primary control actions will increase the generation by a certain amount. In the meanwhile, the loads will decrease somehow due to the fact that most loads (such as motors) are negatively frequency dependant. So, in these control areas, the overgeneration makes actual net interchanges away from schedule, Ni A - NiS > 0. If the frequency bias setting for each control area can exactly match the frequency response in that area, the ACE signal still can be kept as the same as prior to the disturbance, in which case 0, in control areas “A”, “B”, and “D”. However, in control area “C”, the ACE is negative due to the under generation and the AGC action is needed to increase the generation in this area. So with proper frequency bias settings in the system, only the control area where the disturbance occurs responds with AGC action.
M M
A
M
M
B M
C
M
M
M
D
Fig. 2 Interconnected control areas. Although the concept of tie-line bias control is very straightforward, there are practical difficulties with calculation of the frequency bias coefficient. The mathematic equation for β can be formulated as
β
=
1 R
D
(2)
where 1/R is the generation regulation or droop, D is the load damping characteristic From the above equation, it can be seen that the frequency bias is related to the governor and load characteristics. Droop characteristic of governors should be within the range of 3% - 5% ideally. However, the true measured droop characteristic of governors actually range from 6.9% - 12.2% according to a survey conducted by NERC on 177 utilities in 1990 [7]. The characteristic of load is very difficult to measure since loads are changing all the time in a control area.
The solutions to the difficulties with calculation of the frequency bias are the followings:
6
(1) Make the droop characteristic of governors in a control area more uniform. According to the NERC tieline bias standard , each control area should set its frequency bias coefficient as close as practical to the control area frequency response characteristic. The NERC governor guide recommends that generating units 10 MW or greater be equipped with governors for frequency response. Governors should provide a 5% droop characteristic and be fully responsive to frequency deviations exceeding ± 0.036 Hz.
(2) Use averaging techniques to approximate the actual frequency bias. In the implementation of AGC, the averaged frequency bias B is used instead of β. In the NERC policy of “Generation Control and Performance”, several methods of calculation of frequency bias setting are given. Fixed bias setting. A fixed frequency bias value is based on a fixed, straight-line function of tie-
line deviation versus frequency deviation. The fixed value shall be determined by observing and averaging the frequency response characteristic for several disturbances during on-peak hours. Variable bias setting. A variable (linear or non-linear) bias value is based on a variable function
of tie-line deviation to frequency deviation. The variable frequency bias value shall be determined by analyzing frequency response as it varies with factors such as load, generation, governor characteristics, and frequency. Bias and jointly owned generation. Control areas that use dynamic scheduling or pseudo-ties for
jointly owned units must reflect their respective share of the unit governor droop response into their respective frequency bias setting. Fixed schedules for jointly owned units mandate that the Control Area (A) that contains the Jointly owned Unit must incorporate the respective share of the unit governor droop
Jointly Owned Unit
A
response for any Control Areas that have fixed schedules (B and C). The Control Areas that have a fixed schedule (B and C) but do not contain the Jointly owned Unit should not include their share of the governor droop response in their Frequency Bias Setting.
B
C
Minimum bias setting for CONTROL AREAS that serve native L OAD. The CONTROL AREA’S
monthly average FREQUENCY BIAS SETTING must be at least 1% of the CONTROL AREA’S estimated yearly peak demand per 0.1 Hz change. Minimum bias setting for CONTROL AREAS that do not serve native L OAD.
The control
area’s monthly average Frequency Bias Setting must be at least 1% of its estimated maximum generation level in the coming year per 0.1 Hz change. Bias and overlap regulation. A CONTROL AREA that is performing OVERLAP R EGULATION
SERVICE will increase its F REQUENCY BIAS SETTING to match the frequency response of the entire area being controlled. A CONTROL AREA that is performing SUPPLEMENTAL R EGULATION SERVICE shall not change its FREQUENCY BIAS SETTING
7
(3) Incorporate the integral action in the AGC controller so that the ACE will be kept small over the time. III. FILTERING OF ACE SIGNAL In the real world, the ACE signal obtained from the formulation in the section II will potentially be noisy in nature or have spikes, especially in the control area where large non-conforming loads (typically furnace and mill loads) exist. For example, the ACE values often had peak-to-peak swings of 150+ MW in a few minutes due to the numerous non-conforming industrial loads in Northern Indian Public Service Company (NIPSCO) [8]. The similar problem existed in Texas Utilities Electric Company [9]. The noisy ACE signal will cause unnecessary control movements, which will increase the overall generation cost and possibly increase the wear and tear on generation equipment. Sometime, the large load excursion is well beyond the power system’s response capability. In that case, the controlled response of the process is worse than no control at all [10].
Thanks to the response limitation of the governors of the units on regulation, it is impossible to chase the fast changing component in ACE rather than the underlying slow changing trend. Even if it is possible, it is not desirable from economic point of view. So, one goal of the controller is to capture the underlying information in the ACE to send out minimum pulses while maintaining the operation performance. The ACE signal has to be processed before it can be used for decision making. Basically, there are two approaches: (1) to filter the ACE signal; (2) to entitle the AGC the capability of discriminating the longterm excursions from short-term excursions. Sometimes, these two approaches are combined in one controller.
For the filtering approach, there are different techniques, such as PID, moving average. Many techniques developed in the signal processing or conditioning can be borrowed, which are beyond the purpose of this paper. For the approach of differentiating the long-term and short-term excursions, the neural network based method is often used. Sometimes, special logic functions are needed to built in the AGC controller to address the specific problem associated with the characteristic of the ACE signal.
8
Fig. 3 Example of non-conforming loads and the ACE [9]
IV. POST-INSTALLATION/ONLINE TUNING The necessity of post-installation/online tuning comes from several aspects. First, the frequency bias in formulation of ACE needs to be reviewed to ensure the estimated frequency bias is close to the actual one. Especially for the fixed bias setting, the bias coefficient is recalculated annually.
Second, some parameters of the AGC controller need to be tuned or optimized with online data. Some controllers need the historical ACE data.
V. POSSIBLE ISSUES AFTER DEREGULATION The Federal Energy Regulatory Commission (FERC) implemented the intent of the Energy Policy Act in 1996 with Orders 888 and 889, with the stated objective to “remove impediments to competition in wholesale trade and to bring more efficient, lower cost power to the Nation’s electricity customers.” From then on, much research effort has been put on constructing the new structure under the deregulated environment [11], [12], [13], [14], [15].
With evolving of restructuring process in the power utility industry, the picture of new structure becomes more and more clear. In the restructured or deregulated environment, the vertically integrated utilities
9
(VIUs) no longer exist. The new entities include generation companies (GENCOs), transmission companies (TRANSCOs), distribution companies (DISCOs), and independent system operators (ISOs). The ISOs control the ancillary services, which include AGC. Usually, there is only one ISO in a state and it reports to the regional reliability organization and NERC. To some degree, ISO is a much larger control area than presently existing areas.
From the implementation perspective, the AGC can be implemented in hierarchical structure or centralized structure. For the hierarchical AGC scheme, ISO dispatches generation on the system-wide basis and send area requirements to the area control centers, which in turn perform load frequency control and send the dispatch signals to the generation units. For the centralized scheme, ISO performs the load frequency control function on the system-wide basis and sends dispatch instructions to generation units though the area control centers. In this case, the control areas under an ISO no longer provide the AGC service. California ISO took the hierarchical scheme in its initial operation stage. Now it is operating in the centralized scheme. [16]
Although the AGC function is the same in essence under the market-price based new environment, there are some unique features of the AGC at the level of ISO. (1) In addition to the dramatic increase of tie-line transactions in the new open market, the tie-line transactions are much more dynamic in nature. This should be taken into consideration in the design AGC controller to ensure the tie-line transactions are kept on schedule. (2) At the ISO level, the control area is very large. The frequency bias coefficient in the ACE formulation has to be determined carefully to reduce unscheduled interchange. (3) Since TRANSCOs have full freedom to contract with different GENCOs, the AGC system shall be able to track the participation of the generators, again this is also dynamic. (4) The AGC shall have interface with the scheduling application. Along with the progressing of the restructuring, probably more special features of the AGC can be identified under this new environment.
VI. CONCLUSIONS Based on the practical issues around AGC, the design and implementation of AGC should take the different characteristics of the different generation units into account. The advanced algorithm does not necessarily work well in the real-world settings. In addition, the trend toward more deregulated competitive market will definitely impose more implementation issues on the AGC. Although a lot of publications have been addressing the possible technical changes, the final version of the AGC in the open market will be clear only after extensive interaction between the theory and implementation process. It will probably end up with different versions along with different control areas or utilities.
10
References:
[1]
R. L. King, M-L. Ngo, and R. Luck, “Interconnected System Frequency Response,” Proc. of the Twenty-Eighth Southeastern Symposium on System Theory, pp. 306-310, 1996.
[2]
D. B. Eidson, and M. D. Ilic, “Advanced Generation Control with Economic Dispatch,” Proc. of the 34th Conf. on Decision & Control , pp. 3450-3458, 1995.
[3] N. Jaleeli, D. N. Ewart, and L. H. Fink, “Understanding Automatic Generation Control,” IEEE Tran. Power System, Vol. 7, No. 3, pp. 1106 – 1122, Aug. 1992. [4]
O. P. Malik, A. Kumar, and G. S. Hope, “A Load Frequency Control Algorithm Based On a Generalized Approach,” IEEE Tran. Power Systems, Vol. 3, No. 2, pp. 375-382, May 1988.
[5]
M. Azzam, “Robust Automatic Generation Control,” International Conf. on Simulation, pp. 253-258, 1998.
[6] N. Bekhouche, and A. Feliachi, “Decentralized Estimation for the Automatic Generation Control Problem in Power System,” First IEEE Conf. on Control Applications, Vol.2, pp. 626-631, 1992. [7]
“Impacts of Governor Response Changes on the Security of North American Interconnections,” EPRI Final Report, Oct. 1992.
[8]
L. D. Douglas, T. A. Green, and R.l A. Kramer, “New Approaches to the AGC Non-Conforming Load Problem,” IEEE Tran. Power System, Vol. 9, No. 2, pp. 619-628, 1994.
[9]
R. R. Shoults, R. Kelm, and D. Maratukulam, “Improved System AGC Performance with Arc Furnace Steel Mill Loads,” IEEE Tran. Power Systems, Vol. 13, No. 2, pp. 630-635, 1998.
[10] C. W. Ross, T. A. Green, and J. W. Schwartzenberg, “Experimentally Determined Properties of Exponentially Correlated Gaussian Noise,” ISA Transactions, Vol. 13, No. 2, pp. 172-181, 1974. [11] N. Bengiamin, L. Wang, and H. Salehfar, “Assessment of Automatic Generation Control (AGC) in a Deregulated Environment ,” IEEE Power Engineering Society Winter Meeting , Vol. 2, pp. 1331 1336, 2000. [12] B. H. Bakken, and O. S. Grande, “Automatic Generation Control in a Deregulated Power System,” IEEE Tran. Power Systems, Vol. 13, No. 4, pp. 1401-1406, 1998. [13] A. P. S. Meliopoulos, G. J. Cokkinides, and A. G. Bakirtzis, “Load-Frequency Control Service in a Deregulated Environment,” Proc. of 31 st Hawaii International Conf. on System Sciences, Vol. 3 , pp. 24-31, 1998 [14] R. D. Christie, and A. Bose, “Load Frequency Control Issues in Power System Operations after Deregulation,” IEEE Tran. Power Systems, Vol. 11, No. 3, pp. 1191-1200, 1996. [15] J. Kumar, K. H. Ng, and G. Sheblé, “AGC Simulator for Priced-Based Operation, Part I: a Model,” IEEE Tran. Power Systems, Vol. 12, No. 2, pp. 527-532, 1997. [16] F. Albuyeh, and Z. Alaywan, “Implementation of the California independent system operator,” IEEE Int. Conf. on Power Industry Computer Applications, pp. 233-238, July 1999.
11