A Study on Diagnosis of Partial Discharge of Underground Cables Using Deep Learning Diagnostic Model

In this paper, we propose the diagnostic method for partial discharge of underground cables using deep learning. The training data required for the diagnostic model development is collected by the test cells that can generate various PD defect signals. The proposed diagnostic ensemble model was applied to combining MobileNet V1 and statistical model. MobileNet V1 is based on a streamlined architecture that uses depth-wise separable convolution to build light weight deep neural network. The statistical model cans extract the statistical feature values that can represent PD characteristics. As a result of the performance test of the diagnostic model, PD types is classified with 98% accuracy, and the validity of the proposed model was confirmed.