Development of expert systems as on-line operational aids

Abstract

In recent years, the interest in Expert Systems (ES) has grown rapidly. The application of expert systems to power system operations and on-line environments is a new area of research. In this thesis, the issues involved with on-line knowledge based systems for operations is investigated. To begin, the development and validation process involved in building two ES operational aids are discussed. One incorporated logical and heuristic rules for Customer Restoration And Fault Testing (CRAFT) of power lines with automatic switches. The other, referred to as VCES (Voltage Control Expert System), integrated numerical and heuristic techniques for reactive power/voltage control in a power system. Both expert systems are implemented as rule-based forward-chaining systems.Can expert systems such as CRAFT and VCES be applied in an on-line operation environment? The answer depends on the quality of the knowledge base and the computational efficiency. Besides validating the knowledge base, computation time is analyzed. Based on discrimination nets, used in many production system languages for pattern matching efficiency, a method is proposed to obtain the worst case computation time of a rule-based system. The method depends on bounding the individual computational components of the inference engine (pattern matching, conflict resolution, and actions) and partitioning rules into groups performing different tasks in the decision-making. The worst case time is obtained by combining the costs of each of these tasks. This research is a step toward insights into the application of knowledge based systems to power system operations and the efficient implementation of rule-based systems.