Parameter identification and prediction of Jiles-Atherton model for DC-biased transformer using improved shuffled frog leaping algorithm and least square support vector machine

This study presents a novel approach for the modelling of transformer core magnetisation characteristics under DC bias condition by using the inverse Jiles-Atherton (J-A) model. An improved shuffled frog leaping algorithm (SFLA) is proposed to identify the five parameters of J-A model, where an adaptive chaotic mutation operation is added in the global searching process to increase the identification accuracy. With the proposed algorithm, the J-A model parameters under different DC components are identified based on the DC-bias experiment on the real transformer. The conventional SFLA and particle swarm optimisation (PSO) method are also applied to identify the parameters of J-A model. All the identified results are compared with the measured B-H curves to verify their identification accuracy. Moreover, the least square support vector machine (LSSVM) algorithm is used to predict the J-A model parameters of transformer under larger DC component from the previously identified parameters in smaller DC. The calculated results have shown that the improved SFLA has higher identification accuracy than the conventional SFLA and PSO methods. Furthermore, LSSVM algorithm can effectively forecast the transformer magnetisation character under large DC bias condition, which is beneficial for the research of transformer DC bias problem.