Dynamic stability calculations in power system have been mainly analyzed by the time span in the order of second and many researches in this area have been carried out. But today we need long-term calculations in the order of minute because of the complicated bulk system analysis. In Japan, gnerally, dynamic stability is .analyzed by the so-called alternating solution, which uses the 4th order Runge-Kutta method and the Y-method (for circuit calculations) alternatively. This method has high accuracy and is very simple in programming, but in the case of long-term analysis the Runge-Kutta method becomes unstable, and we can not obtain correct solutions. Today, especially in the U. S. A., the trapezoidal method instead of the Runge-Kutta method is widely being used. In the trapezoidal method the differential equations are solved simultaneously with the circuit equations. However this trapezoidal method tends to lead oscillatory trajectories. Because this property is critical for long-term calculations, the acurracy of the trapezoidal method becomes a problem. In this paper a number of numerical analysis methods are tested for finding the best method for the long-term stability calculation, and a suitable scheme for this purpose has been found.
This paper presents a new real-time method for locating fault section at substations. When a fault occurs, a lot of information resulting from protective relay operation and cricuit breaker operation is recorded. Using this information we tried to locate the fault section, especially by considering the sequential relationship of the information and by dividing fault ares. Primarly, in this method, the fault area is divided into several sections based on protection area and operation time of protective relays. Next, expecting subsequent operation of backup relays, “waiting time” for reasoning is given to each divided section. After a lapse of “wating time”, all of the suspected fault sections are extracted and given priority based on empirical knowledge of experts. We developed a prototype of the expert system for fault section location, which was applied to various complicatted fault cases. We proved the effectiveness of our method even in case of multiple faults and no-operation of protective relays and circuit breakers.
Recently, in Japan and abroad, information related to abnormal events occurring in nuclear power plants is being exchanged by utilities and international organizations. The information contains a variety of knowledge which may be useful for prevention of similar events. With this background, an expert system which incorporates the above knowledge into its knowledge base, and offers suggestions about potential abnormal events and preventive know-how, has been developed. The mode of the system utilization mostly recommended by the authors is to infer the potential for abnormal events from newly experienced ones at other plants, evaluate countermeasure priorities, and then consider preventive measures for identified potential events. The system provides six fundamental inference functions for such mode. Among which, the “similar event prediction” function with respect to plant components and the “significance evaluation” function for given event sequences are the main featurings of the system. This paper discusses the system design/construction philosophies focusing on their unique points, presents the inference algorithms and the knowledge data structure going into details of “similar event prediction” and “significance evaluation”, and demonstrates some system operation examples. The knowledge base is now being enhanced in order to put the system into operation in the near future.
In this paper, an expert system was developed for initial stage restoration of EHV power system from system blackout. The operations to be done in case of blackout, are usually determined at every electric station by operation rules, and the power system is restored to the prefault state by carrying out the determined operations at each electric station. In case a regular operation should not work well, several supplementary operations are prepared. This method has some merits that it can restore along normal operation state, that its logic is simple, etc. Because of these merits, a part of the operation rules have been automated by automatic restoration equipments, but in case of system blackout, such human judgements as to secure initial power source are required, and operations by commands from a load-dispatching office become dominant. In this paper, an outline of restoration process was firstly clarified through analysis of the operation rules. Next, the operation rules were transformed into production rules, and were put in order at every electric station. An individual knowledge base was prepared for each group of electric stations in construction of the expert system, and a proper order was introduced in choosing a station to be restored, by which vain search of rules was eliminated. The expert system was applied to a model power system, and it was shown that the computation time for generating all restoration operations is about 0.5s, and that the system can deal with cases where there are some damaged equipments.
To realize an operation supporting expert system for large-scale power networks, two items are considered to be especially important. The first is inference speed and the other is an interface between an expert system and a conventional energy management system (EMS) or a SCADA system. To improve inference speed a dedicated artificial intelligence processor and a specific domain shell for real-time expert systems are adopted, and also reduction of rule-search space size is devised considering features of power networks. Tightly coupled interface and distributed database between a logic-program-based system and a procedural-program-beased system are desirable. An operation supporting system for a large-scale power network has been developed and installed in Fukuoka Integrated Control Center, Kyushu Electric Power Company. The system is configured of triple super-minicomputers with back-ended artificial intelligence processors, and also of dual micro-processor-based front-end-processors (FEPs). Knowledge-based functions, such as fault determination, restoration procedure generation, security monitoring and operation planning are installed in addition to conventional functions of an EMS and a SCADA system. The inference results and their inference speed are satisfactory.