A neural network approach to the detection of incipient faults on power distribution feeders

A neural network strategy for the detection of high-impedance faults on electric power distribution feeders is described. This approach consists of collecting samples of substation current during normal and abnormal feeder operation and using these samples to teach a neural network the rules for fault detection. The learning capability utilized in a neural network approach makes it possible to adapt partially trained fault detectors to individual feeders. The data preprocessing required to set up the training cases and the implementation of the neural network itself are described in detail. the potential of the neural network approach is demonstrated by applying the detection scheme to high-impedance faults simulated on a model distribution system