Anda di halaman 1dari 9

558

IEEE Transactions on Power Systems, Vo1.6, No. 2, May 1991

GUIDING A POWER SYSTEM RESTORATION WITH AN EXPERT SYSTEM


Daniel S. Kirschen Empros Systems International Minneapolis, Minnesota Terry L. Volkmann Northem States Power Company Minneapolis, Minnesota Many utilities have concluded that improvising a restoration plan under the stress of a blackout situation could lead to unacceptable delays. After a careful study of their individual situation, these utilities usually develop one or more written restoration procedures reflecting the most effective ways of restoring the power system [6-121. Although it is commonly agreed that carefully prepared procedures considerably simplify the restoration, these procedures have serious limitations and the development of more powerful methodologies has been suggested [2]. The main disadvantages of a written restoration procedure derive from its static nature. It is most likely that the status of the power system following the blackout will differ from the postulated conditions used to develop the plans. Developing separate procedures for different initial conditions alleviates but does not solve the problem. Difficulties almost always arise during a restoration and it is clearly impossible to prescribe alternatives for every conceivable problem. At some point, the operator in charge of the restoration is thus likely to be forced to stray without guidance from the prepared plan. Such detours can be large or small but their preparation always takes time and their implementation increases the probability of errors. Another disadvantage of written restoration procedures is the amount of time required to reflect modifications in the power system. It is interesting to note that a review of 48 major disturbances lists outdated procedures as the second leading cause of restoration problems [2]. The format of conventional restoration procedures also becomes a hindrance if the procedure cannot be followed as written. Textual information is intended to be read sequentially and does not lend itself to other modes of consultation. Following a written decision tree forces the reader to jump back and forth in the text, querying a written document is absurd, and switching to another level of detail is impossible. Finally, this textual format and the paper medium commonly used to record these procedures force the operator to enter the restoration instructions into the control computer. Transferring information manually is obviously a slow and error-prone process. Sakaguchi and Matsumoto [13] were the first to suggest that an expert system could be used for planning switching operations. Since then this concept has been extended to the restoration of distribution systems [14] and generalized to the creation of optimal switching sequences in substations [15]. Other authors [la. 171 have proposed frameworks for the development of expert systems applicable to the restoration of high voltage networks. Reference [18] shows how an expert system has been used to assist in the automation of the restoration of a small power system. The restoration of a power system is often described as a two-phase process. During the first phase, a skeleton network linking the generating stations and the main substations i s re-energized. This energization is then expanded to the rest of the network and customers are reconnected as power becomes available. This paper describes the Restoration Assistant, a prototype expert system designed to reduce the amount of time required to complete
0885-8950/91/0500-oS58$01.0001991 IEEE

Abstract - Our civilization has become so dependent on a steady supply of electric power that a blackout bears an enormous economic cost. Minimizing the amount of time needed to restore a power system helps reduce this cost and the public's resentment over the interruption of service. An expert system can help the operator achieve this goal by providing actualized restoration plans, dependable recommendations and a clear picture of the situation.
A hierarchical structure has been adopted for the prototype expert system described in this paper. This structure separates the selection of the next objective (strategic reasoning) from the development of a plan showing how this objective can be achieved (tactical reasoning). This separation enables the system to quickly adapt its recommendations when a problem arises during the res toration. Planning connection paths through substations of widely different configurations is greatly facilitated by the adoption of a hierarchical description of the topology of these substations. This knowledge representation scheme is the symbolic equivalent of the one-line diagrams used by the operators. Keywords: power system restoration, power system operations, expert systems, knowledge representation.

Introduction
Although electric utilities constantly strive to maintain the stability and the integrity of their power system, absolute security can never be guaranteed. A catastrophic sequence of events that defeats the preventive measures and overcomes the corrective actions will occasionally unfold and cause a collapse of the entire power system. If such a blackout does occur, electrical service must be restored as soon as possible to minimize the economic loss, the inconvenience to the public and the direct cost to the utility. At the same time, this restoration must proceed carefully to avoid endangering lives, damaging equipment or jeopardizing the progress already accomplished in the re-energization. While performing the numerous operations required to rebuild the power system following a blackout, the operator must be attentive to many unusual voltage, line flow, stability and resource constraints. It is important to note that most operators have probably never had to work with the system when it is so far removed fiom its normal state. References [l] to [5] present very interesting descriptions and discussions of the problems which commonly arise during restorations. 90 SM 341-8 P I P S A paper recommended and approved by the IEEE Power System Engineering Committee-of t h e IEEE power Engineering Society for presentation at the IEEE/PES 1990 Summer Ivieetine;, Xnneapolis, Minnesota, July 15-19, 1990. Manuscript submitted January 31, 1990; made available f o r p r i n t i n g June 21, 1990.

559

h i s the f i s t phase of the restoration of a large power system. T reduction is achieved by:
presenting restoration plans that reflect the actual state of the power system to the operator; relieving the operator from the more mundane tasks of the restoration to let him or her concentrate on more complex activities and on tasks which require human interactions; providing a clear picture of the status of the restoration to facilitate the decision-making process.

human interface. A detailed description of each of these modules is presented in the following sections.

Strategic

Planning

The identification of an internal or external source of power is the first step in restoring a power system. Once this source has been activated, through a blackstart or an agreement with a neighboring utility, it is used to supply the auxiliaries of other power plants and make possible their restart. Backup paths between the energized stations must also be established to reduce the probability of a catastrophic set-back. For the purpose of strategic planning, the expert system decomposes the creation of this skeleton network into a set of objectives. Each of these objectives corresponds to the creation of an energization or backup connection between two stations. While the list of objectives is given, the order in which they are achieved is not imposed. Except for the f i s t two or three steps following the activation of the initial source of power, there are usually several objectives that can be reached directly and often several ways of reaching them. The expert system keeps track of which stations have been energized and which backup paths have been established. The strategic level module also relies on a predefined set of paths which have been determined through engineering studies as being suitable corridors for the restoration. A path is either a single line or a series of lines and minor substations linking two major substations. Once a substation has been energized, the paths that emanate from it become feasible. If such a feasible path leads to a de-energized substation, the energization of this substation becomes a reachable objective. If this objective was already reachable, the newly feasible path is recorded as an alternative way of reaching this objective. Similarly, if a feasible path links two energized substations, it provides a way of creating a backup path between these two stations. A path is not considered to be feasible if it comprises a faulted line. a substation whose communication system has failed, or a substation that has been disabled by a major incident (e.g. a fire.) The predefined paths are not a detailed description of how a connection can be established between two substations. They are defied solely in terms of lines and stations and their sole purpose is to avoid having to consider too many details when searching for a possible approach to an objective. The feasible paths and the corresponding objectives are displayed on a high level diagram of the power network. The operator can thus decide which objective will be undertaken next based on a clear display of the options. If a problem arises in the execution of an objective, the operator can temporarily abandon it, return to the high-level diagram and select a substitute objective. If the restoration is managed from a central location, all the reachable objectives cannot be worked on at the same time and the most urgent one must be selected. The relative urgency of the various objectives depends in part on their intrinsic importance which can be assessed off-line. However, this urgency is also affected by time-dependent factors which can change drastically during the restoration. For example, the priority assigned to the reenergization of a particular station may be high until the expiration of the deadline for a hot restart. On the other hand, the priority given to the supply of auxiliary power to a nuclear plant would increase dramatically if the emergency generators of this plant were to fail. Upon request from the operator, the strategic planning module will reassess the time dependent factors pertaining to each reachable objective and combine them with their static importance to establish a list of priorities. While all

This expert system was developed on the basis of the newly-defined


restoration procedures of Northem States Power Company and may reflect some of this utility's restoration philosophies. In particular, these procedures do not call for the simultaneous creation of several restoration "islands" as is practiced by other utilities. Consequently, the problems related to the existence of separate islands have not yet been studied in detail. However, every effort has been made to keep the architecture flexible and the rules generic to ensure the applicability of this system to a wide range of situations.

Description of the System

In artificial intelligence terminology, the restoration of a power


system is classified as a planning problem. These problems have traditionally been solved by determining a goal state and searching heuristically for a sequence of actions leading from the current state to this goal state [19, 201. This sequence of actions constitutes the plan. If a problem surfaces during the execution of a plan, which is very likely in the case of a restoration, it must be discarded and the time-consuming planning process must be repeated. T h i s approach is thus not practical i n situations where the environment is unpredictable and uncertain. However, the restoration can be decomposed into a set of goals or objectives. Examples of objectives which must be achieved during the f i s t phase of the restoration include the supply of cranking power to each generating station and the creation of backup links to each of these stations. It is then preferable to plan only as far as the completion of the next objective. Using this approach, the amount of planning which must be repeated is much smaller if a plan fails at execution. Furthermore, if the expert system detects that, due to the current conditions, no feasible plan can be found for a given objective, this objective can be temporarily replaced by another. This decomposition also creates a clear and useful separation between strategic and tactical matters. At the strategic level, the expert system suggests to the operator which objective appears to be the most urgent and broadly outlines the different ways of reaching it. At the tactical level, the expert system prepares a detailed plan showing How a given objective can be achieved following a given approach. In the remainder of this paper, strategy thus refers to the decisions which influence the global outcome of the restoration while tactic pertains to the decisions which affect only the success of a particular objective. Once a plan has been reviewed by the operator, the expert system can take care of issuing the commands necessary to implement it. If a problem develops during the execution of these commands, the expert system will notify the operator and suggest an altemative. The expert system consists thus of three modules (strategic planning, tactical planning and execution control) and a graphical

560

the reachable objectives are shown to the operator, the most urgent objective is brought to his or her attention through a special highlighting. The knowledge encoded in the expert system currently is capable of only a rudimentary priority assignment. More knowledge must be acquired to refine these recommendations. The final decision as to which objective to pursue and which path to use rests with the operator and is communicated to the expert system by pointing and clicking on the graphical representation of the objective. The strategic planning module records the selected objective and the chosen path and passes this information to the tactical planning module which is thus assigned the task of establishing a detailed list of actions which must be carried out to achieve this objective. Possible actions include opening and closing breakers, checking the value of a measurement or calling another human operator to request the execution of a particular operation. It must be emphasized that at the strategic level, the expert system reasons only about abstract concepts (such as objectives and tasks) and high level objects (such as stations and paths). Figure 1 illustrates the type of information provided by the strategic planning module. In this example, several objectives can be reached from the substations already energized. Energizing the station BDS is recommended by the expert system as the most urgent objective. However, the operator could instead choose to energize the stations BLL, HBR, ASK or WHT or to create a backup connection between the substations RRK and IVH via the substation CGR.

KCH, WPC, FIS and BRV. Once the energization of BDS is


achieved, the path which was not chosen will be recommended for a backup connection. The path between BYN and BDS is not suggested by the expert system because a fault on the line between LKM and WEF makes it infeasible. Figure 1 is a black and white approximation of one of the dynamic color displays that have been designed for the expert system. On the actual displays, the user can switch to another display, trigger a planning task, or start the execution of a plan simply by pointing and clicking at the appropriate location.

Tactical Planning
The tactical planning module is responsible for translating into a list of actions the tasks defined by the strategic planning module. A task usually prescribes the connection of two stations via a path. A path is either a single line or a sequence of lines and intermediate substations connecting the two stations. Each task is decomposed into subtasks, one for each substation affected by the connection. Each subtask thus prescribes the creation of a connection between two points within a single substation. These two points are: the incoming and outgoing lines for an intermediate substation; the outgoing line and the energized bus closest to this line for an energized station; the incoming line and the auxiliaries bus for a deenergized station. Tactical planning consists therefore primarily of determining the switches that must be opened or closed to connect two points of a subs tation. The knowledge base used for tactical planning is complete if it covers all substation configurations and any pair of points in any substation. Furthermore, the presence of malfunctioning breakers or faulted busses may not prevent it from producing a feasible connection plan if such a plan exists. A tactical plan is correct if it is equivalent to what an experienced operator would do under the same circumstances. In general, this implies avoiding the energization of unrelated lines and minimizing the number of switching operations. However, this minimization is not carried out with a short term perspective. Some extra switching operations are often performed before they are actually needed to facilitate future connections and reduce the number of operations needed in the long run. In addition to the traditional criteria of completeness and correctness, the tactical planning module must also be efficient. It is indeed likely that the expert system would be rejected by the operators if it took more than a few seconds to construct a plan. The development of an expert system which satisfies these requirements has been greatly facilitated by the design of a new knowledge representation scheme. The topology of a substation is commonly defined for algorithmic applications by specifying the components to which each device is directly connected. For expert systems, this approach is not practical because it does not provide a global view of the substation. Planning a connection path based on this information requires a search analogous to finding one's way in a city by relying strictly on the street signs at the end of each block. Such a search is feasible [I51 but very different from the mental process used by an operator to plan a connection path. It is easier and more intuitive to rely on a "symbolic map" of the

XRD GRC

BEN

G-G-0

ROC
= Deenergized =

Substation

Energized Substation Path Substation

e
@
@

----

-=
= Path Line

Line

= Energized Line

= Objective Substation = Most Urgent Objective

11501

Figure 1. An Example of Strategic Information and RecommendationsProvided by the Restoration Assistant.

The expert system suggests two paths for the energization of BDS. The fist goes through IVG and PKN while the second eoes through

561

substation. A hierarchical knowledge representation scheme has been designed to provide this map to the tactical planning module. The main features of this scheme will be explained using the simplified diagram of the RRK substation shown on Figure 2. This station is divided into two levels separated by transformers (RRK345 and RRK-115). Each level consists of busses (e.g. 115-TOP and 115-BOT for RRK-115. 345-TOP and 345-BOT for RRK-345) and branches (A, B,C. E and F for RRK-345,'E.F. G. H, I, J and K for RRK-115). A branch is defined as one, two, or three switches connected in series. Branches belong to different classes depending on how many switches they contain and whether or not they include a transformer. For example, C is a three-switch branch while E is a two-switch-transformerbranch. If both ends of a branch are in the same level, the branch is an intemal branch of this level (e.g. A , B and C for RRK-345, G, H, I, J and K for RRK-345). If the ends are at different levels (e.g. E, F) it is an extemal branch of each level. A branch is also extemal if one of its ends is connected to a line. Several attributes are defiied for each branch. For example, the definition of branch J is: from-end to-end: first-switch first-midpoint: second-switch: second-midpoint: third- switch: 115-TOP 115-BOT J1 E 12 RRKCGR J3 (a bus) (a bus) (a breaker) (a branch) (a breaker) (a line) (a breaker)

The location of the endpoints of each path are then pinpointed using the branches where they are located and the closest busses. Using this information, the type of connection is determined. For example, if the two endpoints are closest to the same bus, a connection will be established along this bus. Otherwise, a more complex connection type is suggested. The rule base is organized in such a way that the most desirable types of connection are considered first. If a path would involve a faulted bus or a breaker that is blocked open, it is rejected and another path is determined. Finally, a list of all the switching operations required to create this path is constructed on the basis of the selected connection type and the status of the substation.

Example 1:
Suppose that it has been decided to energize the station CGR from the station IVH and through the substation RRK. A plan showing how to connect the points labeled IVH and CGR on Figure 3 must be developed. These points being in different levels, the expert system begins by identifying an extemal branch linking these two levels, Branch F is chosen because it links directly the busses closest to the endpoints. The problem is thus decomposed into three subproblems: closing all the switches of branch F. connecting branch F to the line to CGR and connecting branch F to the line to IVH. The first subproblem is trivial. For the second, it is determined that the connection should be made on the bus 115BOT while for the third it should be made on the bus 345-BOT. Figure 3 summarizes the operations required to create this path.

Similarly, the definition of branch E is from-end: to-end: first-switch first-midpoint: second-switch: 345-TOP J El TR-1 E2 (a bus) (a branch) (a breaker) (a transformer) (a disconnect)

Example 2:
Suppose that a connection between the same two endpoints is desired but that it is known that breaker F1 is blocked open. The expert system determines that branch F cannot be used and thus chooses branch E to link the two levels. The line to IVH must therefore be connected to the bus 345-TOP. This path is inferior to the one constructed in the First example because it requires the energization of the line to PRI. Figure 4 summarizes the operations required in substation RRK to create this path. The system also determines which breakers must be opened in station PRI. Figures 3 and 4 are only black-and-white approximations of dynamic color displays. On the actual displays, the operator can modify a plan simply by pointing and clicking on the desired switches. If the planned connection involves a neighboring station, clicking on the name of that station brings up a similar diagram which highlights the operations required in that station. It must be emphasized that the knowledge used to develop the tactical plans is purely generic. It reflects only standard operating procedures and does not include any information specific to a particular substation. Therefore, the knowledge base does not need to be updated each time the configuration of a substation is modified.

The line RRK-CGR has the following definition: from-station: to-station: from-end to-end: RRK CGR J (a station) (a station) (a branch of RRK) (a branch of CGR)

Some operations which must be repeated often when determining a path through a substation become very simple using this scheme. For example, the definition of line RRK-CGR indicates that this line has a terminal at station RRK and that this terminal is on branch J. The definition of branch J shows that this terminal is separated from bus 115-BOT by only one breaker. 115-BOT is thus the "closest bus" to line RRK-CGR at station RRK and the energization of this line will usually require the energization of that bus. This scheme is applicable not only to breaker-and-a-half type substations but also to ring-bus and other types of substations. Since a unified representation of all substations is now available, a single method can be used to plan a connection path between two points in any substation. This method proceeds as follows:

If the two points are in different levels, the problem is first decomposed into the selection of an extemal branch linking the two levels and the planning of a path in each level.

562
TLK AFT RFS

PRI

ASK

MAX

BCK

STY

CGR

RRK-115 Level

RRK-345 Level

Figure 2. Description of the Substation RRK using the Knowledge Representation Scheme of the Restoration Assistant.

TLK

AFT

RFS

PRI

ASK

Figure 3. Tactical Plan for the Condition of Example 1.

TLK

AFT

RFS

PRI

ASK

MAX

BCK

STY

CGR

Figure 4. Tactical Plan for the Condition of Example 2.

e.........vsy

= C o ~ & o n Path = Unaffected breaker

= Breaker to be opened or kept open.

= Breaker to be closed or kept closed. = Breaker blocked in open position.


11502

563

Execution

Control

Conclusion
Reducing the amount of time required to restore a power system following a blackout is the ultimate goal of the expert system described in this paper. Each of the following features of this expert system contribute to this reduction: The plans generated by the expert system reflect the current configuration of the substations, the true status of the power system, and the known equipment problems. Some operations which are usually performed manually can be delegated to the expert system. The expert system not only performs these tasks faster than the operator, but the time saved can be used by the operator for more complex duties or for functions requiring human inter actions. Since the plans generated by the expert system are already in an "electronic" format, the amount of i n i m a l information which must be manually entered is m and the probability of error is considerably reduced. The expert system reasons and interacts with the operator at both the system level and the equipment level. This feature allows the operator to separate minor difficulties from major decisions. Experiments with the prototype system have shown that it is capable of preparing and displaying complete and correct tactical plans much faster than an experienced engineer could point on a substation diagram to the breakers that should be operated. The Restoration Assistant is intended to supplement, not replace, the design of a restoration strategy for each utility. A significant amount of knowledge is acquired from the results of these studies and implicitly incorporated in the expert system under the form of restoration objectives, paths, and priorities. Inserting this knowledge explicitly in the expert system would further increase its power and flexibility.

Once a plan has been reviewed and approved, the operator has the option of letting the expert system take care of its execution. Performing switching operations manually is indeed a slow process. The one-line diagram of each substation must be displayed, the cursor must be positioned on the appropriate switch, the switching command must be issued and confirmed. A substantial amount of time can clearly be saved by automating these manual operations. If the responsibility for implementing the switching commands is delegated to the expert system, the amount of time required to issue an extra command is very small. The expert system can thus be used to execute commands that might be neglected when the switching is performed manually. For example, issuing an "open" command to a breaker which is already open resets the automatic reclosure mechanism which might have been activated by a relay during the system collapse. If this mechanism is not reset, the breaker may close automatically when one of its sides is energized depending on the last breaker operation which took place before the blackout. T h i s closure could reconnect a significant amount of load and cause a major set-back in the restoration. In order to properly schedule the execution of the switching operations, the expert system monitors the completion of the commands which have already been issued. If the event processor of the SCADA system indicates that a command did not produce the expected result, the expert system b r i g s this fact to the attention of the operator and attempts to develop an alternate plan. If the failure is due to a switch blocked in the closed position, a small modification to the plan can usually be found. On the other hand, if the failure is due to a switch blocked in the open position, the entire plan may have to be modified.

Implementation and Tests


The prototype has been implemented as a rule-based expert system on a Symbolics 3675. Its knowledge base was developed using the PROTEUS expert system shell [21] and currently contains approximately 350 rules. A significant portion of the Northern States Power system has been modelled to validate the expert system. Each module was first tested separately. In particular, the tactical planning rules were checked on different substation configurations. A large number of paths were generated within each substation and verified by comparing them to standard operating practices. If the expert system was informed of the unavailability of some devices, it generated alternate paths which were carefully examined and found acceptable. The various modules of the expert system were then integrated to simulate the restoration of the NSP system. If no equipment trouble was introduced, the expert system produced a plan identical or equivalent to the written procedures. Planning the energization of a station along a path which crosses three large substations typically takes about five seconds. A particularly complex connection or the presence of known equipment problems may require a slightly longer time. Planning a restoration path on-line is therefore not only theoretically feasible, but can also be done in a sufficiently short time to be practical and acceptable to an operator.

References
M. M. Adibi et al. "Power System Restoration - A Task Force Report," IEEE Transactions on Power Systems, Vol. PWRS-2, No. 2, pp. 271-277, May 1987. M. M. Adibi et al. "Power System Restoration - The
Second Task Force Report," IEEE Transactions on Power Systems, Vol. PWRS-2, No. 4, pp, 927-933, November 1987.

M.M. Adibi, D. Scheurer, "System Operations Challenges: Iisues and Problems in Power System Restoration," IEEE Transactions on Power Systems, Vol. PWRS-3, No. 1, pp. 123-124, February 1988.
W.A. Johnson et al., "System Restoration - Deploying the Plan," IEEE Transactions on Power Apparatus and Systems, Vol. PAS-101, No.11, pp. 4263-4271, November 1982.

U.G. Knight, "Aids for the Emergency Control of Power


Systems," CIGRE Electra, No. 67. pp. 101-134, December 1979.

R.J. Kafka, D.R. Penders, S.H. Bouchey, M.M. Adibi, "System Restoration Plan Development for a Metropolitan Electric System." IEEE Transactions on Power Apparatus and Systems, Vol. PAS-100, N0.8, pp. 3703-3713. August 1981. R.J. Kafka. D.R. Penders. S.H. Bouchey. M.M. Adibi, "Role of Interactive and Control Computers in the Development of a System Restoration Plan," IEEE Transactions on Power Apparatus and Systems, Vol. PAS-101, No.1. pp. 43-52, January 1982. D. Scheurer, "System Restoration at Philadelphia Electric Company". paper presented at the IEEE PES Workshop on Real-Time Monitoring and Control of Power Systems, Montreal, Canada, October 10-12, 1984.

T. Kojima et al. "Restoration Guidance System for Trunk Line System." IEEE Transactions on Power Systems, Vol. 4, No. 3. pp. 1228-1235, August 1989. I. Takeyasu et al. "An Expert System for Fault Analysis and Restoration of Trunk Line Power Systems." in Proceedings of the First Symposium on Expert Systems Applications to Power Systems, Stockholm-Helsinki. August 22-26, 1988, pp. 8-24 to 8-31. L.R. Blessing, C.K. Bush, S.J. Yak,"Automated Power System Restoration Incorporating Expert System Techniques," in Proceedings of the Second Symposium on Expert Systems Applications to Power Systems, Seattle, July 17-20, 1989. pp. 133-139. R.E. Fikes, N.J. Nilsson, "STRIPS: A New Approach to the Application of Theorem Proving to Problem Solving." Artificial Intelligence, Vol. 2. No. 3/4, pp. 189-208. 1971. E. D.Sacerdoti, "Planning in a Hierarchy of Abstraction Spaces," Artificial Intelligence, Vol. 5, No. 2, pp.115135. 1974. C. Petrie, D. Russinoff, D. Steiner, "PROTEUS: A Default Reasoning Perspective". in Proc. 5th Generation Conference, National Institute of Software, October 1986.

S. Peach, "Hydro-Qutbec System Restoration Synthesis." paper presented at the IEEE PES Workshop on Real-Time Monitoring and Control of Power Systems, Montreal, Canada, October 10-12, 1984.
P.F. Arnold, "Summary of System Restoration Plan for the Pacific Northwest Power System," paper presented to the Power System Restoration Task Force at the 1982 IEEE Winter Power Meeting. E.J. Simburger. F.J. Hubert, "Low Voltage Bulk Power System Restoration Simulation," IEEE Transactions on Power Apparatus and Systems, Vol. PAS-100, No.11, pp. 4479-4484, November 1981. E. Mariani. F. Mastroianni, V. Romano, "Field experience in Reenergization of Electrical Networks from Thermal and Hydro Units," IEEE Transactions on Power Apparatus and Systems, Vol. PAS-103, No.7. pp. 1707-1713, July 1984. T. Sakaguchi, K. Matsumoto, "Development of a Knowledge Based System for Power System Restoration." IEEE Transactions on Power Apparatus and Systems, Vol. PAS-102, No.2, pp. 320-329, February 1983. C.C. Liu, S.J. Lee, S.S. Venkata, "An Expert System Operational Aid for Restoration and Loss Reduction of Distribution Systems, " IEEE Transactions on Power Systems, Vol. 3, No. 2, pp. 619-626, May 1988.

Daniel S. Kirschen received the Electrical and Mechanical Engineer's degree in 1979 from the Universitt Libre de Bruxelles. Belgium. He received the Master's and Ph.D. degrees in Electrical Engineering from the University of Wisconsin-Madison in 1980 and 1985 respectively. He has been working for the EMPROS division of Control Data Corporation since 1985. Terry L. Volkmann received the Bachelor of Science degree in Electrical Engineering from the University of Minnesota in 1978. He has been employed by Northern States Power Company since 1975. He presently is the Superintendent of Operation Coordination with the responsibility of technical support to the System Control Center in the area of transmission operation. In this capacity, he lead the team that developed the knowledge base of NSPs Intelligent Alarm Processor. He has been working on the development of NSP's restoration procedures since 1986.

Z.Z. B a n g , G.S. Hope, O.P. Malik, "A Knowledgebased Approach to Optimize Switching in Substations," IEEE Transactions on Power Delivery, Vol. 5, No. 1,
January 1990.

565

Discussion
H. M. ADIBI

Application of expert systems for assessing the status of distribution systems and determining the appropriate switching operations following major disturbances has been explored in [13,14 & 151 and the authors paper is a significant contribution in this area. The restart and reintegration of bulk power supply however, requires a combination of expert systems and analytical tools so that during the "strategyf9 and lltactical*f planning, the following concerns would be properly addressed [3,16 & 171:
1.

However, during the 10 hours, Scenario I serves 17.4 GWH or 42% more than 12.3 GWH for Scenario 11. Would the authors comment on the effect of change in goal on their strategy planning?
Mega- Watt - Hours Restored 2500

2000

energizing large sections of transmission lines within the acceptable transient and sustained over voltages without the risk of generator self excitation, load pick-up in large increments without the risk of frequency decline and recurrence of outage, maintenance of steady state and transient stabilities as system is being restored and when impedances are large, reduction of standing angles when closing loops to firm up transmission paths, and coordination of thermal plant start-up timings with load pick-ups to bring generators to their stable minimum levels and within the range of major analog controllers.

L
I

2.

.3 * 1500
c-l

g "

2
T i
V

3.

.3 * 1000
500

4.

5.

0 0

9101112

The existing analytical tools are large, comprehensive and inefficient for on-line strategy and tactical planning. Does the authors approach consider the above and other power system constraints and what existing or new analytical tools are used in planning the "skeletont1 network? Under Strategic Planning it is stated that the "strategic level modules" rely on a predefined set of paths which have been determined through engineering studies. This contradicts the statement under Introduction that the "status of the power system following the blackout will differ from the postulated conditions used to develop the plans". Could the authors clarify these two apparently contradictory statements? In order to avoid losing critical times such as hot restart of thermal plants and to take advantage of several sources of black start capabilities within the power system, generally two or more subsystems are restored in parallel. Could the authors clarify Northern State Power's philosophy in adopting sequential restoration? While restoration procedure has a definite beginning, its completion is not so clear cut. Therefore, minimizing the restoration duration as the goal is not as meaningful as maximizing the restoration of service mega-watt-hour as shown in [ 21. The curves are reproduced here for convenience. It can be seen that both scenarios or strategies reach the 1600 Mw load level 10 hours after the start of restoration.

Restoration Duration - Hours Scenario 1

. Scenario 11

M a n u s c r i p t r e c e i v e d J u l y 2 7 , 1990.

D.S. Kirschen, T.L. Volkmann: The interest that Mr. Adibi


shows for our work and his encouraging comments are very gratifying. His remarks highlight some of the areas which we will have to investigate in more details as we continue the development of our expert system. In particular, it is clear that the use of some analytical tools would increase the confidenceof the operator in the recommendations of the Restoration Assistant. At this time, we believe that streamlined versions of the analytical tools will be used to check the plans developed by the expert system. Considering the uncertainty about the value of some parameters and the questionable validity of the traditional models during a restoration, simplified computationsmight be sufficient. The contradiction between the design of the Restoration Assistant and the introductory statement concerning the uncertainty as to the state of the power system after a blackout is only apparent. If this state was totally predictable, prepared plans would be effective and there would be little need for an expert system . If it was totally unpredictable, the expert system would have to have complete flexibility to adapt to any situation and would thus be extremely complex. In practice, the uncertainty is such that the expert system must have some flexibility but does not have to invent a new restoration strategy for each case. It has thus been designed to develop reconnection plans based on a choice between paths which have been shown to be viable alternatives through engineering studies.

566

Although a parallel restoration of two or more subsystems could theoretically shorten the restoration, the implementation of such an approach is difficult when the control of the power system is centralized as is the case for Northem States Power. The operator in charge of the restoration would have to split his or her attention between two areas and it is doubtful that the end result would be beneficial.

Mr. Adibi is entirely correct to point out that restoration Strategies should be compared on the basis of the amount of energy delivered to the customers. It is clear that the choice of restoration objectives should be guided by this criteria.
Manuscript received September 27, 1990.

Anda mungkin juga menyukai