A Convolutional Neural Network-based UHF Partial Discharge Atlas Classification System for GIS

Abstract

In order to solve the difficult problem of partial discharge pattern recognition caused by large amount of partial discharge detection data and complex multi-source, a partial discharge pattern recognition algorithm based on VGG-16 convolution neural network is proposed. The parameters of VGG-16 network model are optimized in convolution layer, pool layer and connection layer by means of migration learning. The VGG-16 model is superior to LeNet-5 model and has higher recognition accuracy.