Identification of Partial Discharge Location in a Power Cable Using Fuzzy Inference System and Probabilistic Neural Networks

Abstract

This article proposes an approach for identification of partial discharge location in a power cable using a fuzzy inference system and probabilistic neural networks. For accurate determination of the partial discharge source, feature extraction of the measured signals is used in the proposed method. White Gaussian noise, which simulates the high noise environment, is added to the partial discharge signals when making measurements. The accurate ratios of the partial discharge occurrence prediction using conventional observations of partial discharge signals via oscilloscopes are much improved by the proposed method. According to the concept of power delivery, both the peak absolute value and average power of partial discharge signals are adopted as input variables of the fuzzy inference system and probabilistic neural networks. Finally, experimental results validate that the proposed approach can effectively determine the location of partial discharge sources in practical partial discharge measurement.