The Application of Wavelet Entropy and BP Neural Network for Fault Location in Distribution System

Abstract

Distribution network (DN) usually has complex structure, numerous branches and small power supply radius. The obvious features lead to accurately locate fault location in DN difficultly. We proposed a hybrid method combing discrete wavelet transform (DWT) and back propagation neural network (BPNN) for fault location in DN. DWT is used to analysis the characteristics of transient signals. Entropy per unit (EPU) indices computed from the DWT decomposition are used as inputs to multi-layer BPNN models serving as fault location predictor. Numerical results based on atypical IEEE 34-bus system are performed to validate the effectiveness of the hybrid method.