Support Vector Machine-based Denoising Technique for Removal of White Noise in Partial Discharge Signal

Abstract

—Partial discharge (PD) measurement has emerged as a dominant investigative tool for condition monitoring of insulation in high voltage equipment. In general, PD signals are severely polluted by several noises like white noise, random noise, discrete spectral interferences (DSI). The challenge lies with removing these noises from PD signal effectively by preserving the signal features. In this article, support vector machine (SVM) based denoising technique has been proposed for the removal of white noise from PD signal. The proposed SVM technique retains the edge of the original signal efficiently and also pseudo Gibbs phenomenon does not exist with SVM technique. In order to evaluate the effectiveness of the proposed method, artificially simulated PD signal mixed with white noise and the measured PD readings are considered. For the purpose of comparison, other denoising techniques such as fast Fourier transform (FFT), discrete wavelet transform (DWT), and translation invariant wavelet transform (TIWT) are also considered. The results reveal that, SVM based denoising technique shows better performance in terms of higher signal to noise ratio, signal reconstruction error ratio, cross correlation coefficient and reduction in noise level, mean square error, and waveform distortion.