Abstract: To improve the accuracy of transformer fault diagnosis, a transformer fault diagnosis method of optimized hybrid kernel extreme learning machine(HKELM)based on improved northern goshawk optimization algorithm(INGO) is proposed. Firstly, the ReliefF algorithm is used to filter and reduce the dimensionality of 19 dimensional transformer fault features. Secondly, three strategies of Logistic tent chaotic mapping, Cauchy mutation operator, and nonlinear increasing weight, are introduced to improve the Northern Hawk optimization algorithm and enhance its global optimization ability. Then, the improved INGO algorithm is used to optimize the initial parameters of HKELM to improve its classification accuracy and robustness. Finally, the features optimized by ReliefF are used as input features for the model and compared with different transformer fault diagnosis models. The simulation results show that the INGO-HKELM fault diagnosis model has higher fault diagnosis accuracy compared to other models.

Key words: transformer, fault diagnosis, northern goshawk optimization algorithm, hybrid kernel extreme learning machine