Research on insulation defect diagnosis method of XLPE AC cable based on partial discharge image feature

Partial discharge (PD) is one of the important characteristics of XLPE AC cable insulation. We designed four defect models in the laboratory, and the sample data are obtained by experiment in order to understand the nature and characteristics of defects in XLPE cable. We further decompose the gray image and then select an appropriate method to extract the features of the image. Then we explore the effect of SVM, KNN, DT and BPNN on the defect pattern in the case of different feature sets. The results show that the SVM classifier with color feature set and the DT classifier with color feature or comprehensive feature set are better when the samples are big enough. And when the sample is small, the SVM classifier with color feature set is better.