Defect Classification in MV Cable Terminations Based on Residual Neural Network

Abstract

This paper presents a method based on residual neural networks to classify poor workmanship defects located on MV cable terminations. A total of two cable groups was used, and each group consists of five different cables with five defects. Each cable has one defect. 120 partial discharge patterns were acquired for each defect (1200 patterns in total). One group of cables was used in the training phase and the other in the testing phase. Residual neural networks, one of the deep learning algorithms, were used in the analysis and classification of the data. The results show that the residual neural network algorithm can be used as an efficient diagnostic tool for classifying poor workmanship defects in cable termination, and partial discharge measurements provide valuable information in classifying defects.