A knowledge-based system for real-time operation planning and event analysis in power system control

The development of computer technology has created possibilities to use artificial intelligence and expert systems in design, control and diagnosis of technical systems. In power systems these technologies can be used in several ways. In energy production, knowledge-based systems can be used for fault diagnosis, event analysis in alarm situations, operation planning and control in abnormal situations, maintenance scheduling and, in complicated systems like nuclear power plants, for plant state analysis and planning of annual maintenance. In energy transmission and distribution, knowledge-based systems can be used also for fault diagnosis, event and state analyses, real time operation planning and control, maintenance scheduling and simulation-based prediction, load shedding, production allocation and restoration planning. Knowledge-based systems can also be used as intelligent user interfaces with a wide range of applications. Our aim was to develop a knowledge-based system for real-time operation planning in a power system control centre. During the study a prototype system called EKA was developed for the power system of the Helsinki Energy Board. The prototype is implemented with object-oriented techniques and it includes a complete model of the 110 kV transmission network and knowledge for the generation and testing of switching sequences. The study demonstrated that the solution method is suitable for the problem. The use of the expert system reduces routine work of the operators and gives them more time to think about the real problems. The major advantages of the system are the easiness of the user interface and the domain-oriented representation of the network. The flexibility of the system also surprised the developers and the users. The use of KEE and Common-Lisp were perfectly suited to the problem. With objects, rules, methods, demons and lisp functions, it was easy to describe the electric network, relations in it, its behavior and the knowledge needed for switching operations. Although the development has been successful on the theoretical side, some severe practical problems exist. First, the current system is too slow to be used for real-time purposes. Second, the lack of editors and documentation methods has caused trouble and much extra work. We hope that, in the future, the problem could be solved with more sophisticated editors and documentation protocols. The next step in the project is to convert the system to be used as a training simulator and test and develop the knowledge base. At the same time the studies of the event analysis will be started. According to our plan, both systems will be installed in the control centre in 1990 or 1991.