A SVM Transformer Fault Diagnosis Method Based on Improved BP Neural Network and Multi-parameter Optimization

SVM multi-class expansion strategy based on BP neural network is used for transformer fault diagnosis, which has higher classification accuracy than traditional multi-class SVM. However, this method needs to train the initial weight threshold to diagnose transformer faults, and its coding calculation process is complicated. To solve this problem, firstly, a new fault diagnosis method for SVM transformer based on improved BP neural network, is proposed in this paper, which embeds the guiding module and increases the weight of the feature vector extracted by the network, so as to improve the diagnosis and classification accuracy of SVM greatly. Then, based on the improved BP-SVM algorithm, a multi-parameter optimization of SVM kernel function and penalty factor based on K-fold cross validation and artificial bee colony algorithm is proposed. By selecting the optimal kernel function parameters, so as to optimize the comprehensive diagnosis method of transformer faults, greatly improve the diagnosis efficiency. Finally, it provides technical support for the inspection of transformer equipment.