You are on page 1of 12

IEEE T a s c i n on Power Systems, Vol. 9 No. 3. August 1994 rnatos .

1399

AN AGC IMPLEMENTATION FOR SYSTEM ISLANDING AND RESTORATION CONDITIONS


Herman B.Ross
Virginia Power Company

Ning Zhu M & IEEE

Jay Giri
Sr. Member IEEE

Barbara Kindel Member IEEE

ESCA Corporation 11120 NE 33rd PI. Bellevue, WA 98004

Even during normal operation, the demands on the AGC function are increasing due to the following trends currently facing the An Island AGC function has been implemented in a large-scale electric utility industry: power system, Virginia Power Company (VP), to assist in system Growth. This means that AGC needs to handle more units emergency and restoration situations. It is the first time that the and tie lines. conventional AGC function has been enhanced to be utilized, rather than to be suspended, under such circumstances. Open transmission access. This is resulting in increased wheeling of power across systems. The Island AGC function is designed to be implemented in both the real-time and the Dispatcher Training Simulator (DTS) Joint ownership of units. This means having to deal with environments. The latterprovides arealisticclosed-loopsimulation dispatch and regulation of units that arejointly owned. This environment that mimics the r a system behavior, offering an el also creates situations where ones load and/or generation is excellent opportunity for operators to be well-trained, aware and physically located in a neighbors power system. prepared for abnormal situations. Increasing numbers of base-loaded units, such as hydro, This implementation has been tested on the VP system model, nuclear, and non-utility generators. This resultsin decreasing whichconsistsof4400busesand775 generators. It hassuccessfully amountsof generation available fr AGC regulationpurposes. o completed custom factory tests, and it is the users opinion that the Integration of power plant computers with AGC. This Island AGC will provide a valuable additional capability for use makes real-time plant data available for use in AGC. during system restoration. Consideration of environmental and emission constraints during dispatch.

ABSTRACT

INTRODUCTION

AGC has been an important, routine, everyday function of realtime power system operations for many decades. Its overall objectives have been to: Match total generation to meet total load. Regulate system electrical frequency deviation to zero. Modify generation to ensure that the net scheduled area interchange is maintained. Allocate generation among the generation sources to minimize operating costs. The first objectiveis typicallyassociatedwith the primary regulation of the governor speed controls; this happens in a few seconds. The next twoobjectivesareattained through thesupp1ementarycontrol action of AGC; this occurs in a time frame of tens of seconds. The fourth objective is met by terbry control using the economic dispatch function within AGC; this occurs in a time frame of minutes.
93 SM 519-9 PWRS A paper recommended and approved by the IEEE Power System Engineering Committee of the IEEE Power Engineering Society for presentation at the IEEE/PES 1993 Summer Meeting, Vancouver, B.C., Canada, July 18-22, 1993. Manuscript submitted Sept. 1, 1992; made available for printing April 21, 1993. PRINTED IN USA

Considerationof unit fuel constraints and plant fuel contracts during dispatch. Impacts of large-non-conforming loads such as steel mills andarc furnaces. These loads swing hundreds of MW in tens of seconds and have a severe impact on AGC. Increased emphasis on AGC performance measures. Increasing need to include transmission security cotistraints during dispatch. Another, more challenging,issue facing utilitiesis the need to cope with system break-up during emergency and restoration Situations. Typically, AGC systems are designed to suspend control during emergency situations. System islanding is one such emergency situation. AGC is specifically designed to immediately suspend control if an islanding situation is detected. AGC detects an island by monitoring system frequency deviation and/or by comparing system frequency with the frequencies of generating units across the system. If system frequency deviates from normal, or deviates from a unit frequency, by greater than some threshold, it is an indication that, potentially, a portion of the system has been islanded and is not an integral part of the interconnection. AGC systems are typically designed to suspend control under such circumstances since AGC controls units to satisfy interconnected system objectives. AGC is suspended primarily to preclude dangerous and incorrect control of units. Under these conditions, some of the units may actually belong to the island so that controlling them is of no avail, in terms of meeting the interconnection obligations of the AGC function. In fact, under

0885-8950/94/$04.00Q 1993 IEEE

1400

these circumstances, controlling them would almost definitely result in worsening the island generatioMOad imbalance and island frequency deviation. This in tum, would result in frequency relay mps of units and loads and eventually lead to a blackout of the island. This paper describesan Island AGC function. It is an enhancement to the conventional AGC function that is intended to allow AGC to be used effectively during emergency and restoration conditions. The Island AGC function discussed in this paper is designed to operate in parallel with the primary AGC function on-line just for the island. This enables an islanded part of the system to run on auto-pilot while the operator focuses on restoring normalcy across the rest of the system. The current implementationrequires theoperatortodeterminethattheislandexistsandwhichgenerators are within the island. The generators in the islanded portion are manually assigned to a second AGC function, which has the primary objective of maintaining constant frequency in the island. This selection is done on-line without adversely affecting the primary AGC function. It is highly likely that the primary AGC function would normally havesuspendedcontrol by now due to the islanding situation. Once the rest of the system is restored, the primary AGC function can be put back on active control. Later,the island can be reconnected to the system and the islanded units put back in the primary AGC; thereby reverting to the normal case of a single AGC function.

THE MOTIVATION

2.1 Why Island AGC? Utilities are continuing to focus on system breakup emergency scenarios and to address the means to cope with them effectively. As part of their system planning objectives, many utilities are designing their protection schemes to automatically create viable islands during cascading emergencies rather than blacking out portions of the system. The guideline being used is that a successfulisland is a much better condition that a blackout. Thus, systems are being designed to survive cascading emergencies by automatically creating viable islands that have a reasonably good generationbad balance. This ensures that. although the power system is operating in an abnormal degraded state, customers are continuing to be served. Once this condition is reached, the operational challenge is to ensure a continual balance between generation and load and to maintain the island frequency a OHz. t This is where the Island AGC function fits in.

Provision of a second AGC function, which maintains frequency of an island for the short-term, will help the operator focus on issuing manual directives to restore the rest of the system. This feature can be a valuable asset to the operator during this period of restoration. The operator can essentially ignore the island while he addressestherest of the system. Later this island can be reconnected to the rest of the system. When the reconnection is to be made, the The Island AGC function can be implemented in the real-time and Island AGC will be set to control to the same frequency schedule in the DTS environments. By implementing this function in the as the primary AGC function. This will allow the islands phase DTS, the operators have the opportunity to experience this angle to come in phase with the rest of the system in a smoother emergency situation in a secure, off-line environment without manner, significantly reducing the bump that might otherwise affecting real-time operation. They can study, at their own pace. occur, and helping to insure that the connection will remain intact. the behavior of power flows and frequencies in the system and their dynamic characteristics during system breakup, islanding and 2.2 The Virginia Power System subsequent reconnection of the island. This permits a better understanding of the power system phenomena that occur during TheVirginiaPowersystemisgeographically dispersed with several such emergency conditions and, in turn, provides confidence in major generation and load centers. Some portions of the restoring the system efficiently. By practicing alternate courses of aansmission system are more heavily loaded than others. The actionsinalessstressful,lessintimi&tingandsecureenvironment, Virginia Power system is along the east coast of the United States operators are better prepared to confidently handle emergency wheredamageby windstorm and thunderstormsmay causetripping of heavily loaded transmission lines. Under certain circumstances, situations if and when, they arise in real life. this might result in additional cascading trips, possibly creating an The major functions and features of the Island AGC function, as island within the Virginia Power control areas. Although an well as test results on the large-scale VP system model, are islanding situation has not been experienced in reallife, engineering presented in this paper. studies indicate that, during emergency conditions, the potential for automatic creation of islands within the V r i i Power system igna exists.

3 THE CONVENTIONAL AGC FUNCTION


The AGC function has played an important role for many years in power system control, monitoring, and operation. A functional diagram, along with a short description of the interfaces and major functions for the conventional AGC system, follow:

1401

3.1 AGC Model

Figure 1. Block Diagram of the Conventional AGC System

4
3.2 The Interfaces The AGC function interfaces with the following EMS functions:

THE ISLAND AGC FUNCTION

4.1 Ovenrlew

SCADA State Estimator Security Enhancement Load Forecast Real-Time Operating Plan
3.3 The Major Functions The major functions of the Conventional AGC System are:

The followingdiagram (Figure2a.) showsa high-level configuxation of the Island AGC function.

W
Figure 2a. Island AGC Configuration The typical sequence of actions in invoking the Island AGC function would be: The islanded situation is detected by the State Estimator through network topology processing.

Load Frequency Control (LFC) Economic Basepoint Calculation (Economic Dispatch) Reserve Monitor AGC Performance Monitor

1402

Once the islandis identified, the operator manually specifies the frequency source to be used as well as the units to put under the Island AGC. In the fume implementation, this step eventually could be pedormed by the software, which could automaticallyassign the available fresuencysource and uniuwithintheislandedareatotheIslandAGC,accordingto the islanding information from the State Estimator. After activating the Island AGC, the operator can manually adjust the control parameters (such as frequency bias or scheduled frequency) in order t obtain appropriate dynamic o performance and to speed up the restoration. Eventually, after the rest of the system has been restored to normalcy and the frequencies in both the primary and the island areas are close enough, the island would be reunited with the primary area.
An easy-to-use user interface provides a flexible tool for the operator to modify the Island AGC control parameters. The on/off status of the Island AGC, the desired island frequency source and units, as well as the Island AGC parameters, are all controllable by

4.2.2

Software Design Because the islanded situation is not a normal operating condition, the Island AGC function itself is different from the conventional AGC function in the following ways: 4.2.2.1 The Control Mode The island is isolated from the rest of the system, which means no tie lines exist. The Constant Frequency Control (CFC) mode is used instead of the typical Tie Line Bias Control (TLBC) mode. Also, AGC is specified to be always working in emergency status instead of being dependent on the value of ACE to select the mode (normal, assist, or emergency). 4.2.2.2 The Frequency Bias In the past, numerous investigations have been conducted to learn how to specify the frequency bias and to select the proper settings [4-61. Two different ways are provided in this Island AGC implementation. Thefirst isthesameastheoneusedin theconventional AGC function: enter a constant value. The second way is D enter an island frequency bias as a function of the current island operating status. This feature brings the bias closer t the system's nature and o offers more flexibility in improving the dynamic performance.
4.2.2.3

the system operator through these displays. Figure 2b is a hard copy of the Island State display which shows the current operating information as well as the major AGC parameters of the islanded

The Impacts

area.

The impacts of the Island AGC on the original AGC software are: Plant Controller (PLC). Each PLC is under the control of either the Primary AGC of the Island AGC, depending on its designation, and the PLCs in each area are responsible only for the regulations in their own area. Therefore the PLC regulation factors, as well as the PLC control signals, should be calculated separately to correspondto its designated area. Dynamic Interchange. An island is an isolated area with no d o n s to any neighbaing areas; it will not participie in dynamicinterchangesuntil it is reconnected with theprimary area
Cum*13.20

Losser

Figure 2b. Island AGC Display


4.2 The Island AGC Features
4.2.1

Reserves. The units in the island can no longer contribute to the reserves of the primary area. They should be excluded from both the operatingand the spinningreserve calculations in the primary area. 4.2.3 Operator's Options The Island AGC function designed here offers operators control and freedom on the following issues: SWStop the Island AGC function. The system operator has the control of the time and location to start the Island AGC function. The operator also determines when the island is to be integrated with the primary area and when to stop the Island AGC function. Define/Adjust control parameters. The major parameters, such as the frequency bias and schedules in the Island AGC function,can be entered and adjusted by the system operator through an easy-to-use user interface. Different values can be used during different phases to adapt and improve the system dynamics. Specify the island. The system operator has thgfull freedom to choose which island to put under the Island AGC control. He can select the frequency source and the units for the specified island. This option gives the system operatormore flexibilityin handling the system restomion under an islanded situation.

Working Environment

The main purpose of the Island AGC function is to help with system restoration under the islanded situation. Therefore, the function is designed to supplement real-time control and to run simultaneously with the primary AGC function. Although the islandingsituation o c rarely in the ra world, it can cm el causesevereanddamagetothepowersystem. Inorderto train system operams to deal with such an unusual situarion with greater confidence, the Island AGC function is designed to be implemented in the DTS environment This offers an excellent opportunity for system openuon to simulate the islanding scenario on their own system. They can observe al the events in sequence under l different system operating conditions, study the feasible preventive andrestorationsaate~es,manipulateoHo@rimaryandisland)parallel AGCs, and study the effect of adjusting parameters -as a whole-to prepare for possible islanded situations.

1403

ALEXANDRIA-ARLINGTON

F A I R F A X AREA

Lancaster
White Stsne

Figure 3. One-line Diagram of the Northeast Portion

1404

TESTS AND RESULTS

6 om 4.

5.1 TestScenarlo

5 omo.
4w.m-

The tests were performed, under the DTS environment, on the Virginia Power system model with 4400buses, 5000 lines, 2133 loads and 775 generators. The islanding potential exists in several portions of the system, the northeastportion, shown in Figure 3, was chosen here as our test bed. The test began with several lines in the north comdor, which is an imporranttransmissionlinkbetween thenortheastand thenorthwest, being tripped. LINE 2024 between the stations LANEXA and CHICKHOMINY. -LINE 285 between the stations WALLER and CHICKHOMINY. LINE 92 between the stations PROVIDENCE FORGE and LANEXA. This event could occur during a severe windstorm or as a result of the overloaded situation of these lines. The redistribution of the power previously transmitted through these tripped lines might cause overload on the other lines. As a result, the area could be islanded by the consecutive CB openings at the following lines: LINE2016between thestationsLANEXAandHARMONY VILLAGE. =LINE 85 between the stations SHACKELFORDS and HARMONY VILLAGE. LINE 224 between the stations DUNNSVILLE and LANEXA. LINE214betweenthestationsSURRYandWINCHESTER. .LINE 263 between the stations NEWPORT NEWS and WINCHESTER. The northeast islanded area usually appears as a load in the VirginiaPower system. At the moment the system was islanded in the test, the primary system had 10961 MW generation, 10540 M W load and 421 MW interchange. The sudden loss of load caused an over-generation situation in the primary area, so the generatorstended to speedup (only slightly), while the situation on the island wasjust the opposite. The mismatch between generation and load created an under-generation case. With a much smaller inertia, the frequency deviation in the island is much bigger than that in the primary area, as observed from Figure 4.

m.m200.001m.m-

-4o.m

t (mW 60.rnJ...................--r--.--.--.----.--...........,...........................,........................... 1 5 10 15 20
1.""

Figure 5. The ACE Trends During The Islanded Situation


5.2 Test Results Figure 4 displays the frequency profile before, during and after the islanded situation. Figure 5 shows the ACE curves in both the primary and the island areas during the islanded period.

It is apparent, from studying Figure 4, that the frequency of the primary area was hardly disturbed because of the much greater inertia and much smaller relative load disturbance; while a rather big ACE at the islanded moment appears in Figure 5, which was caused by the sudden line tripping. (The realdelivered interchange jumped from 421 M W to 980 MW at the islanded moment.) The islanded area suffered much more than the primary area. The island frequency had a rather big dip at the beginning because of the severeunder-generation conditionand the much smallerinertia. The advantage of having a smaller inertial is the quicker dynamic response, while the disadvantage is the possible overshoot. These features are displayed in Figure 4. There is always a trade-off between the speed of the response and the amount of the overshoot. In the islanded frequency restoration t process, the bigger the bias factor, the faster the response but a the same time, the larger the overshoot. This can be observed from Figure 6 where curve A uses a smaller bias factor (5%) while curve B uses a greater one (10%). The dynamic performance canbe improved greatly and restoration achieved in less time with the assistance of experienced system operators. Taking the system security and performance into account, control parameters can be adjusted on-line and some restrictions can be relaxed for certain situations (all based on the current operating conditions). For example, the over-generation situation in the primary area during the islanded period can be drastically relieved by manually shutting down some generating units in the Bath County, especially at the beginning stage. The quicker the reduction of the mismatch between the load and generation will greatly improve the system performance so that the ACE in the primary area decreases much faster,as Seen from curve B in Figure 7. On the other hand, the under-generation condition in the islanded area can also be improved through the operator's assistance. The recovery of the island frequency wouldbe smoother

.......M-y
--RimUy-y

59.912

59.907 59.800 59.600


3 '

,I('

.........*' ...'...*.

,,..'.
59.400 59200-

..*e

,..'...'
5

,,*10 15 20
PR(UI0II

t(mln)

Figure 4. The Frequency Trends During The Islanded Situation

1405

and faster if the basepoint settings of the islanded units are monitored closely and adjusted properly, especially when the frequencydependent load takes a fairly large percentage. In addition, putting some fast response units in the manual conml mode and ramping up/down at their maximum ramp rate is also helpful. This effect can be observed from Figure 6 by comparing curve C with curve A, where curve C involves the operator's manual control of the islanded units in Yorktown station besides the Island AGC while curve A does not. Because the two AGC functions are independent, the disturbance of controls only affect the area where they occur. This can be seen by compglingFigure5 andfigure7 wheretheACEcurveintheprimary ~ v a r i e s a ~ * g todifferentoperatingconditions, ~ t l y ~ while the i l n ACE cutve is not i f u n e at all. sad nlecd The system was integrated when the frequencies in the primary and the island areas were close enough for a smooth reunion, all the CBs opened before were closed and the Island AGC was turned off to allow the Primary AGC function to take care of the whole integrated system.

6 CONCLUSIONS In the past few years, the power industry has been focusing more
closely on system restoration issues. The Island AGC implementation described here, presents a novel apptoach for AGC to be used during the islanding condition and system restoration. After islanding, the operator's attention is simultaneously drawn to many diverse and critical issues. This new AGC implementation allows the operator to quickly create a second, short-tern AGC function just for the island. "he islanded portion of the system can then operatein a self-sustaining,automatic mode so that the operator can focus attention on restoring the rest of the system. At the reunion of the island, the electrical aberration that occurs at the moment of reconnection will be significantly reduced and may prove to be almost "bumpless." This will help insure that the reconnection will hold the fvst time rather than having the interconnect line@)relay open, which may then expose the island to enough instability that it could shut down entirely.

The Island AGC function can be implemented in both the real-time and the DTS environments. The DTS provides a realistic closedloop simulation environment which mimics the real system t o behavior. It is an excellent environment for development, demonstration and training. The authors are aware that the Island AGC function is not field proven and that it may not be readily provable until a real situation occurs and someone "steps up to the bar" to try it out. We do, however, believe thatan idea such asthis must be tested analytically viasimulationasafmtstepindetenniningtheconceptualfeasibility of the process. The ESCA simulator is a mature product. The simulator used extensively in the paper to test and validate other real-time EMS functions (such as AGC and the StateEstimator) in an offline secureenvironment, has shown that manually initiating Figure 6. TheIslandFrequency Trends Under Different Conditions Flat Frequency Control (after the island is identified) of the islanded area can maintain the island frequency at the scheduled level. ACE

"I
500.00

: U
200.00
100.00

i"

The results from the Virginia Power system's simulated islanding demonstrate a successful implementation of the Island AGC function as a supplement to the primary AGC function.

-20.00

-004 6,0
1

...........

.............. .............
5

.............. ..............
10

I '

............. .............

..... t(mln)
20

15

Figure 7. The ACE Trends Under Different Conditions

1406

The

d e g the and Daniel Monda for their editorial assistance.

Ph.D. from Clarkson University in New York, in 1977, ACKNOWLEDGMENTS for his support where he was also a Visiting Assistant Professor. Dr. Giri to thank Chuck Ellis is an active member of the IEEE Power Engineering Thanksalso go to Lindsay Society and Eta Kappa Nu.

BARBARA E. KINDEL received the B.S. degree in psychology from the University of Wyoming in 1983 and B.S. and M.S. degrees in electrical engineeringfrom New [l] L. K. KirchmaYer, Control of Mexico State University in 1986 and 1987, respectively. Systems. New York, NY: Wiley, 1959. She is currently a Lead Engineer at ESCA Corporation in Bellevue, Washington. Ms. Kindel is a member of Eta [2] IEEE Current Operational Problems Working Group, "CurrentOperatingProblemsAsswiatedwith Automatic Kappa Nu, Pi Mu Epsilon, and Tau Beta Pi. rn. Generation Conwl," IEEE T a s PAS-98, No.l.1979. [31 T. M. Athay, "GenerationSchedulingand Control,"Roc. Of IEEE, Dec. 1987, pp. 1592-1606.

REFERENCES

141

N Jaleeli, et al., "Understanding Automatic Generation .


rn. Conwl," IEEE T a s on Power Systems, Aug. 1992, pp. 1106-11 12.

[5] L. S. VanSlyck, et al., "Implications of Frequency Bias Settings on Interconnected System Operation and Inadvertent Energy Accounting," IEEE TPS, M a y 1989, pp. 712-723.
[61 T. Kennedy, et al., "Variable, Non-Linear Tie-Line FrequencyBias for InterconnectedSystem Control,"IEEE

TPS,Aug. 1988. pp. 1244-1253.


[71 North American Electric Reliability Council Operating Manual, Guide I. Systems Conwl, Dec., 1987.

BIOGRAPHIES
HERMAN ROSS received his BSEE in 1965 from the University of South Carolina and has been serving Virginia Power Company since then. He has served in District Operations, District Engineeringand System Power Supply Departments. His primary service and tenure has been with Power Supply in engineering and supervisory roles supportingthe System Operation Center (SOC) computer systems. In addition, Mr. Ross has served and is currently sewing as the Project Manager for the Energy Management System (EMS) which is being supplied to Virginia Power by ESCA Corporation. NlNG ZHU received her M.S. from Nanjing Automation Research Institute (NARI), China in 1983 and her Ph.D. from Arizona State University in 1991. Dr. Zhu worked for NARI as a System Engineer (1983-1987) and has been serving ESCA Corporation as a Senior Power System Engineer since April 1991, working on Generation/DTS/ Network areas.
JAY GIRI, Ph.D., is currently a Principal Engineer who

leads the power systems applicationsteam, and a cofounder of ESCA Corporation, Bellevue, WA, USA. Dr. Giri is also an affiliate associate professor at the University of Washington in Seattle. Dr. Giri completed undergraduate studies at the Indian Institute of Technology in Madras, India in 1969; he received his

1407
Discussion
M. L. OATTS (Southern Company Services, Inc., Birmingham, AL. 35202): The authors have presented, in a n easy to follow paper, an interesting
scheme for providing an Island AGC application. Upon review, several implementation issues come to mind. The authors thoughts on these would be greatly appreciated. First, the premise for utilizing the proposed Island AGC implementation assumes the existence of what the authors call a "viable" island. Is this implementation only intended for use with islands created via controlled methods such as the protection schemes referred to? If not, several questions come to mind

help he can get. It is encouraging to see that this anomaly is now being addressed. The authors articulate clearly the need for such capability, and recognize the steps that are required: detection of the islanded situation; identification of the equipment comprising the island; adjustment of control parameters; re-integration of the system. At present, these steps are performed manually, but the potential for automation is pointed out by the authors. Also, and necessarily, this new AGC function has been tested only by simulation using an dispatcher training simulator. Several aspects of the newly designed function remain unclear. First, it is unclear whether the authors are addressing intentional or unintentional islanding. They cite the development of automatic creation of islands during cascading emergencies as a means of avoiding total blackout of the system. (A prominent example of this development is the "defence plan" now being perfected by Electricit6 de France [A].) Both intentional and unintentional islanding occur during rapidly deteriorating situations. In the case of intentional islanding, coordinated underfrequency load shedding may be employed to establish a rough load-generation balance. In the case of unintentional islanding, the loadgeneration imbalance within an island is likely to be large, and any control, to be effective, must be imposed promptly and emphatically. While it is not clearly stated in the paper, the overall impression given is that the authors are dealing with unintentional islanding. In either case, but especially in the latter, it is likely that the present reliance on manual implementation will be unsuccessful, since the rapidity with which such situations evolve will not permit of timely and effective operator action. Thus, effective deployment of the new AGC function will require that it be fully automated throughout the first three steps: island detection, island definition, and parameter adjustment. The second point that remains unclear is how the frequency bias is calculated "as a function of the current island operating status." Presumably, this is done in some simple, straightforward nianner, but several such formulae can be postulated, having various degrees of plausibility. A description of the formula actually used would contribute to assessment of the new function. The authors are congratulated on presentation of a pioneering concept. [A] C.Counan et al.: Major Incidents on the French Electric System - Potentiality and Curative Measures; IEEEiPES Summer Meeting, 1992, Seattle (92SM432-5PWRS)

* How are uncontrolled (or randomly) created islands presented to the operator so he quickly understands its makeup and can utilize the
Island AGC (e.g., select a frequency source, define units in the island, etc.)? Is a special display used, does he rely on island identification on buses from a power flow, other? This is often one of the most difficult problems in an island situation - operator communication.

* If the State Estimator is used as indicated in the paper to notify the


operator of an island, is t h i s necessarily sufficient for determining an island? The situation which comes to mind is the continued connection of sub-transmission level lines to the "island identified by the State Estimator. Often, only bulk power transmission networks are modeled by the State Estimator and in the case mentioned a frequency island would not actually exist.

* How are sufficient frequency source points to be selected and defined in the database to cover the "random" island condition? Off-line
simulation of island creation, frequency source at all plants, other? If instead, the Island AGC is intended for use only with the controlled creation of an island, other questions arise. Hovr would the creation of such an island be ensured by the protection scheme mentioned? How would such a scheme consider the actual dispatch settings of the units and actual "island load when determining "a reasonably good generatiodoad balance" as the paper describes? The utilization of unit capabilities is obviously not sufficient since the actual loading may not meet the needs of the island when it is created. For example, units in the island area upon its creation are on minimum but the island load requires near maximum unit output. Since most AGC controllable units can not normally move instantaneously to the desired output of the island, what would the protection scheme do? Unless some intelligent, adaptive protection scheme is utilized, it would seem that a rigid, preset tripping scheme to create the island could cause problems in the island (such a s generator trips or load shedding) for system conditions different from that for which the scheme was developed. Any test results and operator feedback which you have obtained with the Dispatcher Training Simulator that address the preceding questiondissues would be very interesting. Finally, some general philosophical questions come to mind concerning the idea of using Island AGC. The authors stated that one benefit of the implementation is that it facilitates the phase matching of the island to the rest of the system to reduce the "bump" which would occur upon reconnecting the network. This benefit is understood. Does this not, however, have to be balanced against the risk involved in doing "normal" AGC in the island? Since the determination of the frequency bias factor to be used by the operator in the island will be difficult at best in the "random" island condition, why not let the plants in the island hold frequency via manual control? This is especially true if a "guess" at the frequency bias is used. It does not seem unreasonable to expect that the plant can probably do control as well as an "untuned AGC. The movement of plant output by plant control would seem to be a less risky alternative under a very tenuous system condition, especially considering that the need to tightly hold a scheduled island frequency is questionable. Several parts of the world control their system frequencies manually and this would seem appropriate in this case. The control of the frequency via plant intervention would seem to minimize the risk of the Island AGC program overcontrolling units or deciding to move them to inappropriate loading points under abnormal system conditions. Would the authors please comment?
Manuscript received August 12, 1993.

Manuscript received August 16, 1993.

RICHARD P. SCHULZ, American Electric Power Service Corporation, Columbus, Ohio 4321 5: The authors are to be commended on having put a basically good idea into practice in their System Operating Center (SOC). Tomas DyLiacco has described his experience with use of a local supplementary control within power plants, that is put in service where the plant is islanded [I]. Those applications differ in that they are in smaller, developing utilities, while Virginia Power (VP) operates as part of a large interconnection with an experience record of no islanded operation with local frequency control in recent decades. The experience of several upsets [2 51 suggests several things about the nature and formation of electrical islands within larger interconnections: 1) usually there is a generation-load imbalance within the island immediately upon its formation; indeed, the power flows across the boundary of the to-be-formed island are usually a factor in the island formation. Then there is consequent rapid change in the island frequency immediately after its formation, with frequency rates up t o *3 Hzlsec; 2) the

...

LESTER H. FINK, ECC, Inc., Fairfax, Virginia. Suspension of


automatic control ("tripping to manual") during emergency conditions has always seemed somewhat incongruous, since those are precisely the conditions under which the operator needs all the

1408

boundaries for the island are unpredictable in interconnected power systems that rarely experience islanding operations; 3) Where operators are experienced in analyzing in handling one situation, they tend to use the lessons learned in that experience in other situations in which they may not be appropriate. These observations suggest these questions about the operator training on Island AGC with the dispatcher training simulator (DTS): Does the dispatcher training simulator have the a) capability of simulating any of several scenarios with different island boundaries and different load/generation imbalances within the island? Can the trainer who directs the DTS choose the b) timing and circumstances of the island formation? These features will enhance the training of operators by including training in the ability to recognize and act on unexpected conditions.

Union Elec. Co., Illinois Power Co., "Islanding of Union Electric Co. and Western Part of Illinois Power Co. Systems, 10:32 a.m., February 13, 1978", Report to MAIN, April 24, 1978. EPRI, "Simulators: Tough Training for Top Operators", EPRI Journal Vol. 13, No. 4, June, 1988, pp. 22-29. Schulz, R.P., "Capabilities of System Simulation Tools for Analyzing Severe Upsets", Proceedings, International Symposium on Power System Stability, Ames, Iowa; May, 1985, pp. 209-21 5.
Manuscript received August 16, 1993.

H. B. Ross, Jr.:
It also has been observed [61 that power system simulations are unlikely t o capture the response of the islanded part of the system, in large part because of poor knowledge about the response of controls and protective devices in these unusual operating conditions. Have the authors considered changing the modeling parameters used within the DTS from time to time so that the operators will experience the unpredictability of an islanded system's response to the islanding? This would enhance the ability of the operator t o respond t o unexpected consequences of islanding, including complete collapse of the island.
The authors' presentation of the paper described a mechanism within the automatic generation control (AGC) software for using topological network analysis t o determine (by software) the boundaries of an island, and to communicate the boundaries to the operator when islandb) occur. They also described the use of several frequency sensors throughout the VP area to provide an independent indication of island boundaries should an island form. This suggests some additional questions. Does the topological islanding detector directly suspend AGC action to the units within the island? Have the authors considered using phasor measuring devices to indicate the islanding of an area? There have been circumstances in which Eastern American Interconnection utilities have created islands during normal switching operations; these islands were characterized by slowly diverging frequencies. The phase displacement from within and outside the island would be a more immediate island indicator than would frequency in those conditions. Again, the authors are thanked for having described this uncommon improvement in the ability to control the power system during unusual circumstances. Personal Communication, Tomas E. DyLiacco, July 21, 1993. Davidson, D.R., Ewart, D.N., and Kirchmayer, L.K., "Long Term Dynamics Response of Power Systems: An Analysis of Major Disturbances", IEEE PAS-94, MayIJune 1975, pp. 819-826. Brand, C.W., Usry, R.O., "Gulf Coast Area System Disturbance Final Report", Southern Services, Inc., Birmingham, AL, October 26, 1973. Vol. I, Text and Various Data; Vol. 2, Appendices.
The authors wish to point out that an error was found, after the formal paper had been submitted, on page 6 in section 5.1; the last bullet item should read:
Line 263 between the stations NEWPORT NEWS and CHUCKATUCK.

This error was commented on a t the presentation of the paper. The authors apologize for this oversight and ask for the reader's indulgence.

Author's Discussion : The authors wish to thank all those who attended the formal presentation of the paper and a special thanks to those who prepared written discussions. In retrospect, the paper probably should have been titled in such a way that the restoration process would have received the main emphasis rather than the system islanding process. We say this because, indeed, the restoration concerns were the primary motivation for developing the Island AGC function. Under the restoration scenario one is dealing with an already sharply defined geographic area that is without electric power and the load pickup on the available generation is being carefully added manually by field personnel as they reconnect circuits within the island area. With this situation definition, several of the questions voiced by the discussers would probably not have been of concern. While the primary goals of this fiinctionality are to aid in the system restoration processes and to serve as a training tool for the electric power system operator, by simulation of islanded areas, it is hoped that the functionality will in time come to fruition and do all that is alluded to in the paper and the various discussions. With this in mind, the questions raised by the discussers are addressed in the following paragraphs. For purposes of clarity we have listed the questions raised by each discusser and followed it with our response.

Discussion

by:

Mr. Richard P. Schulz American Electric Power Service Corporation Columbus, Ohio 43215-2352

1409
onestion 1 :

Does the dispatcher training simulator (DTS) have the capability of simulating any of several scenarios with different island boundaries and different load/generation imbalances within the island?
Answer 1: Yes, the simulator is capable of simulating most real world operational configurations. This is so because the scenario is built by using devices in the DTS' real world model of the transmission and sub-transmission equipment, lines, breakers, switches, generators, etc. Whatever island boundaries the trainer or engineer wish to establish, taking into account the real world electrical capabilities, can he set up for the scenario to run and the capability even exists to interactively change the scenario as the simulation progresses.

indicate the islanding of an area because our emphasis was not on islanding occurring in the real-time situation but rather what do we do to smooth the restoration process once an island has occurred or whenever the electric system is being pieced back together after a large scale blackout. As such, we have no plans t o develop instrumentation schemes that would manage transient conditions during a system break up event.

Diecussion bu: Mr. Michael L. Oatts Southern Company Services, Inc. Birmingham, Alabama Ouestion 6: Is this implementation only intended for use with islands created via controlled methods such as the protection schemes referred to? Answer 6: No, there is no "controlled creation" of the island capability. Some utilities do design protection schemes this way. The traditional protection schemes are the definitive measures in responding to power system events. Question 'z: How are uncontrolled (or randomly) created islands presented to the operator so he quickly understands its makeup and can utilize the Island AGC.

The load/generation imbalances are not modelled in great detail. Rather, the island area loads will be adjusted during the scenario by a simple load curve model for the load(s). The amount of imbalance can be controlled by modifying the load curve data.

9 2 -:
Can the trainer who directs the DTS choose the timing and circumstances of the island formation?
Answer 2: Yes, please refer to Answer 1 the first paragraph. The trainer can designate specific equipment and the times each item or group of equipment is switched open or closed by the simulator's scenario process.
Question 3:

Answer:
The topology processor will identify an island condition by messaging the electric power system operator. This messaging together with the (EMS) computer actuated circuit breaker indicators on the large mapboard within the System Operation Center provide the system operator with a presentation scheme that will help him to determine that an island exists and where its boundaries lie.
Question 8: If the State Estimator program is used as indicated in the paper, to notify the operator of an island, is this necessarily sufficient for determining an island? h s w e r 8: Not in many cases due to the possibility of connection of lines at the sub-transmission level that are not monitored by telemetry. However, if the State Estimator program does message the operator that an island exists, the operator should then closely review his mapboard indications and the sub-transmission oneline diagrams shown via his workstation CRT. The operator can also monitor the various frequency transducer readings, from around the control area, for a significant difference with the transducer that is within the general area of question. Virginia Power has installed frequency transducers in five selected power station sites that are located throughout its service area.
Ouestion 9:

Have the authors considered changing the modelling parameters used within the DTS from time to time so that the operators will experience the unpredictability of an islanded system's response to the islanding?
Answer 3: Yes, we have considered this issue. The DTS can simulate not only deterministic events but also conditional events as well as probabilistic events. The modeling parameters, such as the load model and its characteristics, the generator responses, the relay models, probabilities of occurrence, etc., can be changed easily within the DTS.
Question 4:

Does the topological islanding detector directly suspend AGC action to the units within the island?
Answer 4: If the question is; does the detector suspend AGC to only those units within the island, the answer is no. When an island is first detected, AGC is "tripped" (stopped) for all units in the control area (both inside and outside the islanded area). Additionally, there is program logic for unit frequency checking which will suspend AGC to a unit if that units' frequency differs from the electric power system frequency by more than some threshold since this would indicate that the unit is probably in its own island. Ouestion 5: Have the authors considered using phasor measuring devices to indicate the islanding of an area? Answer 5: We have not considered using phasor measuring devices to

Does not the benefit of Island AGC to reduce the electrical transition ("bump") of reconnecting the island into the network have to be balanced against the risk involved in doing "normal" AGC in the island?

GoSWer.
The concern expressed in this question is shared by the authors. For this reason the implementation provides the operator with

1410

a means to, (1)observe the manner in which the island reacts to the AGC actions delivered to the generators within the island so that he can (2) manually adjust the AGC frequency error bias factor to dampen the control action if it is causing too much oscillation of the island area's frequency. One of the purposes of having the island AGC function is to "free" the operators from being occupied by manual duties when they have so many critical tasks to do during the islanded situation. I t is hoped that simulator environment training will help the operators and the engineering support staff to develop some practical guidelines for using this capability in real-world situations. Diseuseion by: Mr. Lester H. Fink, ECC, Inc. 4400 Fair Lakes Court Oakton, Virginia 22033

the electric power system operator. The authors agree that there are many concerns to be dealt with if this concept is to be applied to a dynamic event wherein an island is being created.

Would the authors provide the formula for the automatic calculation of the frequency bias that the functionality would recommend to be used?

AmYfdk
The programming would accept each unit that the operator directs as being in the islanded area and compute the value we refer to as the "calculated island frequency bias" (CIFB) as: CIFB = K SUM ( LFCMX of the units under the Island AGC )

where K is enterableladjustable by the computer system programming analyst. The recommended value is:

tion 1Q: Are the authors addressing intentional or unintentional islanding?


Bgswer 1 : Q Please refer above to the opening discussion where we say that the primary intended uses for this functionality are in the electric power system restoration process and in the training of

K=

____________________----------SUM ( LFCMX of the units under the Primary Area AGC )

Primary Area Frequency Bias

Manuscript received September 24, 1993.

You might also like