关键词: 支持向量机（SVM）, 参数优化, 人工鱼群算法（AFSA）, 变异, 变压器故障诊断, 决策模型

Abstract: In this paper, a new method for fault diagnosis of power transformers is proposed based on Support Vector Machine（SVM） and improved Artificial Fish Swarm Algorithm（AFSA）. Aiming at the problems of blindness of search, slow rate of convergence, and low accuracy of optimum solution, an improved AFSA is presented to solve the optimization problem of SVM. Firstly, Cauchy mutation is adopted to improve the prey behavior of the artificial fish. Secondly, based on the characteristics of [t]-distribution and information searched by fish swarm, death and renascence mechanism is adopted for inferior fish to enhance the capability of artificial fish survival and evolution, which can improve the efficiency and accuracy of the algorithm. The improved AFSA is employed to optimize the parameters for SVM. The result shows that the convergence of the improved ASFA is relatively faster and much more precise than that of the classical one and the performance of SVM classifier is improved at a certain extent. Finally, Decision Directed Acyclic Graph（DDAG） is adopted to extend SVM for settling the multiclass classification problem and a decision model for fault diagnosis of power transformer is established based on DDAG. Moreover, compared with Grid-SVM, GA-SVM, and PSO-SVM models, the results demonstrate the higher diagnostic accuracy based upon the proposed approach and show that the proposed model can be used as an effective tool for fault diagnosis of power transformers.

Key words: Support Vector Machine（SVM）, parameter optimization, Artificial Fish Swarm Algorithm（AFSA）, mutation, transformer fault diagnostic, decision model