A Method for Transformer Fault Diagnosis Based on IWOA-BPNN

In order to improve the accuracy of transformer fault diagnosis, an improved whale optimization algorithm (IWOA) combined with back propagation neural network (BPNN) is proposed in this paper. Firstly, since the initial population is generated randomly, and the quality of the initial population has a direct impact on the performance of the algorithm, chaotic mapping (CM) is used to initialize the population of the whale optimization algorithm (WOA), which is conducive to expanding the search scope and finding the optimal solution. Secondly, inertia weight is a key parameter in WOA, and a fixed weight will lead to a decrease in computational efficiency, which is not conducive to global optimization. The larger the inertia weight, the easier it is to get the global optimal solution. The smaller the inertia weight, the stronger the local optimization ability. The introduction of adaptive inertia weight (AIW) in WOA can improve the global optimization ability and avoid the local optimization. Finally, the weights and thresholds of BPNN are updated by using IWOA, and the BPNN model with optimized parameter values is obtained and applied to transformer fault diagnosis. The experimental results show that IWOA-BPNN model has higher diagnostic accuracy and faster iteration speed than BPNN and WOA-BPNN model, and can effectively diagnose transformer faults.