Cable Fault Detection and Diagnosis Method Based on Convolutional Neural Network

Abstract

In this paper, a method based on convolutional neural network is proposed for the accurate identification of early faults in cables. The method can accurately classify and identify faults from overcurrent signals including constant impedance faults, inrush currents, and capacitive switching disturbances. First, the features of the overcurrent signal are extracted by wavelet transform. Then, construct a convolutional neural network, adjust network parameters through training, and establish a mapping relationship between input features and category codes. In order to solve the problems of training overfitting and learning efficiency, this paper modifies the loss function and uses the method of adaptive learning rate to optimize the convolutional neural network. Through simulation experiments, the effectiveness of the proposed method is verified, indicating that the method can effectively classify overcurrent signals, accurately identify early cable faults, and has high engineering application value. The research results of this method are of great significance for eliminating hidden dangers of cable faults in time.