Artificial neural network based method for fault location in distribution systems

This paper describes a method for fault location in power distribution systems based on artificial neural networks. The topological and electrical characteristics of typical distribution feeders are appropriately taken into account. The input-output relationship of the problem has been studied in detail and, a modular architecture has been derived in order to improve the generalization capability of the employed neural networks and to reduce the magnitude of their training sets. Promising results were obtained for cases of substantially high fault resistance values, on which analytical methods tend to fail. An application tool is being implemented with a windows-like graphic interface, allowing the user to choose among different types of neural network