Automatic suppression of white noise and repetitive pulses interference in on-site partial discharge monitoring

The on-site partial discharge (PD) monitoring is the most powerful method to detect insulation faults in electrical power components, but it is often severely affected by external interferences. This paper designs two algorithms for automatic detection and suppression of white noise interferences (WNI) and repetitive pulse interferences (RPI) in PD monitoring separately. The first algorithm for WNI suppression is based upon wavelet transform (WT), and it solves the issue of mother wavelet selection, decomposition level determination and threshold selection during WT denoising. The second one makes full use of the differences between PD pulses and RPI in pulse waveform and phase distribution, it takes three key steps including identifying pulses, clustering the similar pulses and analyzing pulse-phase histogram to separate RPI. Both of the two algorithms can be used in on-site PD monitoring, and their performance with practical data is shown in this paper.