A new connectionist expert system scheme for distribution system fault diagnosis

This paper proposes a new connectionist expert system (CES) scheme for distribution system fault diagnosis by effectively combining expert system (ES) and artificial neural network (ANN) technologies. Successful test results on a real life distribution system, located within the Taipei West District Office of Taiwan Power Company, derived from the application of this scheme are also presented. When presented with distribution system trouble tickets, the proposed CES analyzes these tickets for fault conditions, based upon human heuristic reasoning encoded in the knowledge base of the ES component, and provides recommended corrective actions to the operator. The CES is characterized by three functional modules. The ANN module, which runs in real time, receives diagnostic messages from a trouble ticket and classifies the failure status and component status, The ES module receives the component status, performs the expert analysis function and provides recommended actions to system operators. The operation interface (OI) module, which is built in a window-based environment, provides the system operator with a user-friendly interface to the developed system. The performance of the proposed CES is evaluated, both with respect to accuracy and running time, in comparison with two other existing solution methods, namely, the fuzzy logic-based expert system and discrete event simulation