Probabilistic neural networks for power line fault classification

Abstract

This paper presents a new power line fault classification scheme using a probabilistic neural network (PNN). One of the major features of PNN stems from its modular architecture design and can be easily extended to adapt to a changing environment by incremental learning. Another distinguishing advantage of PNN comes from its fast training speed as compared to backpropagation. An explicit confidence measure can also be obtained which directly supports the decision made by the PNN. Preliminary experimental classification results of various AC power system faults and transients indicate that the PNN is suitable for power line fault classification.