基于GA与PSO混合优化FCM聚类的变压器故障诊断  

Transformer fault diagnosis based on optimized FCM clustering by hybrid GA and PSO

作　　者：雷浩辖[1] 刘念[1] 崔东君[1] 马铁军[1] 徐海霸[1] 

机构地区：[1]四川大学电气信息学院,四川成都610065

出　　处：《电力系统保护与控制》2011年第22期52-56,共5页Power System Protection and Control

摘　　要：针对FCM聚类、GA-FCM聚类以及PSO-FCM聚类在进行变压器故障诊断时的不足,采用了GA与PSO混合优化FCM(GAPSO-FCM)聚类来进行故障诊断。GAPSO-FCM聚类进行的是全局搜索,克服了FCM聚类容易陷入局部极小值的问题。GAPSO-FCM聚类是以全局最优个体将GA聚类与PSO聚类有机地联系在一起,GA与PSO共用一个最优个体,迭代过程中既包括了GA运算也包括了PSO运算。它依据GA的随机性扩大了搜索范围,之后在所找到的个体附近依据PSO进行更细致的搜索,克服了仅基于单一GA或PSO优化的FCM聚类的早熟问题。通过仿真与实例分析,表明采用GAPSO-FCM聚类进行故障诊断的正确率比采用其他三种聚类的正确率高。Optimized FCM clustering by hybrid GA and PSO （ GAPSO-FCM ） is introduced to diagnose the fault of transformer in order to conquer the shortages of FCM clustering, GA-FCM clustering and PSO-FCM clustering in transformer fault diagnosis. GAPSO-FCM clustering carries out global search, conquering the problem of FCM clustering easily falling into local minimum. According to the best global individual, GAPSO-FCM clustering makes GA algorithm and PSO algorithm organically link together, GA and PSO share a best individual, and the iterative process includes GA operation and PSO operation. It enlarges search area by the randomicity of GA, then searches more carefully according to PSO roand the founded individual, conquering the premature problem of optimized FCM clustering based only on single GA or PSO. Simulation and case analysis indicate that GAPSO-FCM clustering for fault diagnosis is of higher accuracy than the other three kinds of clustering.

关 键 词：变压器 故障诊断 遗传算法 粒子群优化 模糊C均值聚类 

分 类 号：TM407[电气工程—电器]