The method of power capacitor partial discharge signal extraction based on several algorithms fusion

Partial discharge (PD) is the initial sign of power capacitor insulation failure, and an effective extraction and analysis method of PD signal in power capacitor can improve the effectiveness of the on-line fault monitoring. The method which based on wavelet threshold denoising and mathematical morphology alternate mixing filter is put forward in this paper. First, the PD signal which includes noise is wavelet decomposed with the proper wavelet function and scale, and each layer of wavelet coefficients is extracted. Then appropriate flat structure element is chosen aimed at the scales of wavelet coefficients, and the mathematical morphology alternate mixing filter is used to filter out the noise and get a new wavelet coefficients. Choosing the appropriate threshold rules for the new wavelet coefficients, the wavelet coefficients after the threshold treatment is wavelet reconstructed and the PD signal after denoising is got. The wave distortion mean square error e_MSE and SNR are chosen as the indicators to analyze and compare the denoising effects of wavelet threshold denoising method and the fusion denoising method. The results show that the fusion method can effectively remove the noise and keep the original shape characteristic of the signal better.