NOVEL METHOD FOR HIGH IMPEDANCE FAULT RECOGNITION BASED ON SIGNAL PROCESSING AND NEURAL NETWORK

Abstract

A novel hybrid method is presented in this paper, which involves the combination of radial basis function neural network RBFNN with discrete wavelet transform DWT. The proposed hybrid method specifically detects and distinguishes high-impedance fault HIF from other transient events, such as capacitor and load switching. A signal has been extracted using wavelet transform to acquire updated and accurate information from the current signal during the fault. The RBFNN classifier has been presented and used to detect and classify HIF from normal conditions to improve the protection scheme in terms of accuracy and computational time. The new approach provides a robust characterization and classification of different fault conditions in terms of a variety of fault resistance and fault location. The hybrid approach has been examined by performing extensive simulation studies, and the outputs are compared with the previous state of the art, which clearly shows the significance of the proposed method.