A Noise Suppression Method For Partial Discharge Based On Genetic Algorithm

Power transformer is an important electric power equipment in power system,but the power transformer is usually affected by the partial discharge of power transformer,leading to the degradation and aging phenomenon of power transformer insulation,which seriously threatens the safe operation of the power transformer.When the partial discharge of the power transformer occurs,the ultra-high frequency electromagnetic waves would radiate outward from the partial discharge point.Through the ultra-high frequency(UHF)online monitoring technology,the partial discharge of power transformer can be effectively monitored.This article mainly focuses on the ultra-high frequency(UHF)the key technology of on-line monitoring technology,namely,partial discharge signals filtering processing,as the research object.Based on the understanding of mechanism and characteristic parameters of partial discharge of power transformer,the generation and propagation characteristics of the partial discharge UHF electromagnetic wave in power transformer is analyzed and a typical UHF on-line monitoring of partial discharge is presented.The sources of noise affecting the partial discharge are analyzed.The filtering principle and operation process of the continuous wavelet transform method and the FFT threshold denoising method is introduced to in filter the partial discharge signals,and the filtering effect of the two kinds of filter method is analyzed by the simulation analysis.In addition,a noise suppression method for partial discharge based on genetic algorithm is proposed in this papre,and the implementation process and simulation example are given out.The filtering effect of continuous wavelet transform is related with some factors,such as wavelet basis function,the level of wavelet decomposition and threshold processing.The reasonable selection of these factors is crucial to achieve good filtering effect of continuous wavelet transform.The continuous wavelet transform method can only effectively filter the white noises mixed in partial discharge signals.The filtering effect of FFT threshold denoising method is related with the power spectrum threshold coefficient,and the appropriate power spectrum threshold coefficient can make the FFT threshold denoising method achieve the best filtering effect,but FFT threshold denoising method can only effectively filter the periodic narrowband noise mixed in partial discharge signals.The proposed noise suppression method based on genetic algorithm is carried out upon the advantage and disadvantage of the continuous wavelet transform and FFT threshold denoising method.The propoed method not only effectively surpress the white noises as well as the periodic narrowband noise.Moreover,the denoising parameters can be obtained automatically,improving the denosing efficiency.