Acoustic Detection and Decision Fusion Recognition of PD in Power Cable

Acoustic detection has features of non-contact monitoring, no influence on the operation of tested equipment, and little interference by electromagnetic noise. In this paper, the acoustic method is applied to detect the partial discharge (PD) in power cable. Meanwhile, the accuracy of recognition is improved by signal processing, multi-feature construction, and algorithm optimization. The short-term energy and centroid frequency of power spectral density is extracted and the acoustic PRPD is constructed. A decision algorithm based on the fusion of improved K-Nearest Neighbor (KNN) and Back Propagation Neural Network (BPNN) is proposed. Cables with three types of defects were manually processed and acoustic PD samples were acquired. The results show that the PD can be detected quickly by KNN, and acoustic PRPD features used by BPNN showed a high recognition rate. The reliability of PD recognition is effectively improved by combining with the results of KNN and BPNN.