A Comparative Study of Power Distribution Over voltages Classification Using feed forward Artificial Neural Networks and General Regression Neural Networks

Abstract

Power outages due to overvoltage and following damages affect a large part of power distribution networks. A right identification system for the overvoltage can help in fast intervention and shorten the outage duration. In the present work, authors present a comparative study of distribution overvoltage identification. The study is conducted with data simulated from ATP-EMTP with a network designed from data of real case studies conducted on 10kV distribution network of the city of Qingyuan electrified by the China Southern Power Grid. The comparison study is conducted with data from time domain analysis and Wavelet Packet Decomposition (WPD) analysis. Two identification tools are used: the General Regression Neural Network (GRNN) and the Feed forward Artificial Neural Network (FANN). Training and identification performances are compared at the end of the study bringing out some specificities of the two identification tools regarding identifications of the direct strike lightning overvoltages, temporary overvoltages and capacitor bank energization overvoltages.