Fault Diagnosis Study of Shipboard Zonal Distribution Power System Based on Convolutional Neural Network Driven by Feature Fusion Strategy

Traditional convolutional neural network (CNN) in the process of fault diagnosis feature extraction, the information transfer between the layers of the information loss problem, for this problem this paper on the CNN all-connected layer implementation of feature fusion, to solve the problem of information loss. Firstly, a simulation model of the shipboard zonal distribution power system is constructed using MATLAB/Simulink simulation software, and the working conditions of the system short-circuit faults are simulated; secondly, ten kinds of short-circuit faults in different locations of the system are simulated, and the fault data are collected; then the CNN fault diagnosis model is designed using the TensorFlow framework; Finally, the network model is trained and improved to obtain the optimal diagnosis results of the shipboard zonal distribution power system. The diagnosis results show that the CNN driven by feature fusion strategy has good diagnosis effect.