Research on Wavelet Neural Network for Fault Location in Power Distribution Network

Abstract

The distribution network own characteristics such as: complicated topology structure and variable operation mode, so the fault location problem of the distribution network has not been solved. Based on the analysis of the single-phase ground of the system, this paper combines wavelet analysis and neural network. The method not only utilizes time–frequency features and zoom capabilities of wavelet analysis in order to pick up the characteristics of component, but also take advantage of the ability of non-linear fitting of neural network. These studies establish mapping between the characteristic component and fault location, thereby, resolving successfully the problem of the precision positioning, the thought of fault location has fully theoretical basis. The data of EMTP/ATP simulation shows that the scheme has high precision, adaptable and reliable performance.