Hybrid fault analysis system using neural networks and artificial intelligence

The system detects the fault type and approximate fault points using information about operated relays, circuit breakers, and the fault voltage/current waveform. Faulted sections are estimated in the expert system (ES) part and fault voltage/current waveform recognition is performed in the NN part. Since power systems require high reliability, the system uses a verification procedure for the result of waveform recognition obtained by NNs, based on model-based reasoning (MBR). Four different types of NNs were examined and an appropriate NN was selected for waveform recognition. This system is a combination of NN, ES, and MBR technologies for analyzing faults, and it provides functions which cannot be obtained by conventional methods.<
>