Optimization of SVM transformer fault diagnosis by multi-strategy improved Grey Wolf optimization algorithm

Abstract

In order to improve the diagnosis accuracy of transformer fault model, a fault diagnosis method based on MIGWO support vector machine (MIGWO-SVM) optimized by multi-strategy improved grey Wolf optimization algorithm was proposed. Firstly, the Gray Wolf optimization algorithm (GWO) was improved by using nonlinear dynamic weight, fusion arithmetic optimization search strategy to update position and cosine perturbation strategy for the defects of poor global exploration ability, not easy to converge and low accuracy of the original Gray Wolf algorithm. Then, compared with GWO, PSO, SFO, MIGWO algorithm on Sphere and Griewank test function, the experiment proves that the improved multi-strategy improved gray Wolf algorithm is better than the comparison algorithm in both convergence accuracy and search speed. Finally, MIGWO algorithm is used to optimize the hyperparameters of SVM. Perform transformer fault diagnosis. he results show that the diagnosis accuracy of the proposed model is 91.67%, which is 13.33%, 8.33% and 5% higher than that of PSO-SVM, GWO-SVM and SFO-SVM, respectively. It is verified that the proposed method can effectively improve the performance of transformer fault diagnosis.