Feature extraction and classification of partial discharge signal in GIS based on Hilbert transform

Abstract

In order to effectively extract the features of partial discharge (PD) signals in GIS, a method of pulse waveform feature extraction based on Hilbert transform was proposed. The PD data of four typical insulation defects were collected by using GIS PD test platform built in the laboratory. The envelope of PD signal was obtained by Hilbert transform, and the statistical characteristic parameters, time domain characteristic parameters and frequency domain characteristic parameters of PD pulse waveform were extracted based on the envelope. Kernel principal component analysis (KPCA) was used to reduce the dimension of PD feature parameter, and random forest classifier was used to classify and identify PD feature parameters after dimensionality reduction. The identification results show that the proposed method can extract the features of PD signals effectively, and the extracted feature vectors can correctly identify PD caused by different insulation defects.