Classification of Partial Discharge Images within DC XLPE Cables in Contourlet Domain

This paper presents a new method to improve the insulation defect classification rate of DC cross linked polyethylene (XLPE) cables. A partial discharge (PD) test platform is built in the laboratory and four DC cable defect models are designed. The PD signal is collected by a high frequency current transducer (HFCT) and the q-Delta t-n image is constructed. The contourlet transform is performed on the PD image, and the Tsallis entropy feature of each subband coefficient is calculated. The error-correcting output codes (ECOC) classifier is optimized by the adaptive cuckoo search (ACS) algorithm optimization sparse random (SR) coding matrix to realize the defect classification. The experimental data show that the ACS-SR-ECOC classifier in this paper has shown promising performance in the classification of each defect and the Gaussian noise model. Compared with to four kinds of traditional ECOC classifiers, the proposed method has higher overall classification accuracy, which provides a new approach for the defect type identification of PD image in the DC cable.