Fault analysis system using neural networks and artificial intelligence

The authors propose a hybrid fault analysis system using an expert system (ES), neural networks (NNs), and a conventional fault analysis package (CFAP). The system detects fault type and approximate fault points using information from operated relays, circuit breakers (CBs), and fault voltage/current waveforms. Faulted sections are estimated by ES and the fault voltage/current waveform is analyzed by NNs. Since power systems require high reliability, the system uses a verification procedure based on CFAP for the result of NN waveform recognition. Four different types of NNs are compared and an appropriate NN is selected for waveform recognition. With NNs, ES and CFAP used together, the system can obtain the convenient features of these methods.