Classification of Partial Discharge Images within DC XLPE Cables Based on Convolutional Deep Belief Network

The classification of partial discharge is of significance to diagnose the defects in high voltage cable systems. To improve the insulation defect classification accuracy at DC cross linked polyethylene(XLPE) cables, a new method used for PD images classification based on convolutional deep belief network (CDBN) is proposed in this paper. Firstly, four kinds of defects in XLPE cables are designed and tested under DC voltage. The q-Δt-n image is constructed based on PD signal collected by HFCT. Then the diagnostic CDBN model is constructed to extract the high-level detailed feature of q-Δt-n images with Gaussian visible units. Finally, classification experiments with CDBN, deep belief network(DBN), support vector machine(SVM) and backpropagation neural network(BPNN) is conducted. The experiment results show that the proposed method has higher classification accuracy of insulation defect diagnosis.