Study on PD detection for GIS based on autocorrelation coefficient and similar Wavelet soft threshold

To address the issue of white noise at partial discharge (PD) ultra high frequency (UHF) signal in gas insulated substation (GIS), this paper develops an external sensor and proposes new empirical mode decomposition (EMD) denoising method based on autocorrelation coefficient and similar wavelet soft threshold. Four types of typical GIS defects at the PD UHF signal were obtained through experiment. The autocorrelation coefficient of intrinsic mode functions (IMF) components at the PD UHF signal was computed, the cut-off point between the noise signal dominant mode and the UHF signal dominant mode was found. The similar wavelet soft threshold denoising was performed on the signal which is dominated by the noise signal, all the UHF signals were finally reconstructed. The signal-to-noise ratio computed by the proposed denoising method was compared with the one computed by the wavelet denoising method, the results shows that the proposed denoising method in this paper is more effective than the wavelet denoising method.