DC cable feature extraction based on the PD image in the non-subsampled contourlet transform domain

A new method to extract the features of direct current (DC) cross linked polyethylene (XLPE) cable partial discharge (PD) image based on non-subsampled contourlet transform (NSCT) is proposed in this paper. Four types of PD images are obtained from the artificially designed cable insulation defects. The features include Shannon entropy, Renyi entropy, Euclidean distance, and Mahalanobis distance of the optimized subband coefficients by NSCT. The optimized subband coefficients are selected by the immune algorithm optimized affinity propagation clustering method. The back propagating neural network, k-nearest neighbor, support vector machine and decision tree algorithms are used to classify the extracted features. The experimental results show that the proposed method can be applied to the feature extraction of PD signals in DC XLPE cables. Compared with the wavelet method, the recognition rate of the method is improved by more than 16.80% under the same classifier. It is found that the type of defect within insulation can be classified efficiently with the features extracted by the proposed method with a reasonable degree of accuracy.