Application of Convolutional Neural Network Transfer Learning in Partial Discharge Pattern Recognition

Abstract

Due to the lack of expert experience and the shortcomings of high blindness, the traditional partial discharge feature extraction method of gas insulated switchgear (GIS) has an impact on the accuracy of pattern recognition; convolutional neural network emerged in recent years has the ability to adaptively extract features, but training a network with better performance needs to increase the network depth on the one hand, and more supportive data on the other. Therefore, this paper propose a GIS partial discharge pattern recognition method based on transfer learning of three pre-trained network models (VGG, InceptionV3, and Resnet50) under the small data set. And the feature extracted by the network are applied to SVM classifier which performs well on a small data set. Realizing the combination of deep learning and traditional machine learning. Experimental result shows that this method can effectively improve the accuracy of GIS partial discharge pattern recognition.