Fault Diagnosis of High Voltage Circuit Breaker Based on KPCA and PNN

Abstract

To promote fault diagnosis accuracy and efficiency of high voltage circuit breaker, a novel fault diagnosis approaches based on kernel principal component analysis (KPCA) and probabilistic neural network (PNN) is proposed in this paper. Firstly, eight critical features are extracted from tripping/closing coil current curve. Then, KPCA is applied to eliminate redundancy among features, and small scale, more informative and mutual orthogonal components are extracted. Finally, crucial principal components are used to establish inputs and the novel fault diagnosis based on PNN is proposed and studied. The results obtained by testing practical samples reveal that the proposed KPCA-PNN fault diagnosis model can reduce redundancy and improve fault diagnosis efficiency and accuracy effectively, which has good application prospect.