Detection and de-noising of PD signal contaminated with stochastic pulse interference using maximal overlap discrete wavelet transform

Abstract

Detecting partial discharge signals (PD) emanated due to defects and faults in cable insulation is crucial to cable condition diagnosis. However, PD detection proves challenging as the detected signal comes with superimposed noises with various spectrums. One of these common noises that are often present on site is stochastic pulse interference (SPI). This type of noise is random in nature, yet it displays a similar shape to the partial discharge generated pulse and, hence, it becomes difficult to remove (de-noise), while maintaining the original signal intact. In this paper SPI noise is mathematically modeled and is combined with the PD signal for signal processing. The maximal overlap wavelet transform (MODWT) is then used to de-noise the signal using various parameters. Over a wide range of noise magnitude, the parameters that offer best de-noising are subsequently identified. Comparing the MODWT technique with the empirical Bayesian WT (EBWT) proves it to be superior.