AI优化模糊核聚类算法的变压器DGA分析  

Transformer DGA Analysis Based on Fuzzy Kernel Clustering Algorithm Optimized by Artificial Immune

作　　者：宋志杰[1] 王健[1] 

机构地区：[1]华南理工大学电力学院,广州510640

出　　处：《电力系统及其自动化学报》2014年第4期39-43,共5页Proceedings of the CSU-EPSA

摘　　要：针对模糊聚类及核聚类算法在电力变压器DGA分析中存在的初值敏感及易陷入局部极值点的问题,提出了一种人工免疫优化模糊核聚类的新算法。该算法将基于克隆选择原理和亲和力成熟的免疫克隆算法与模糊核聚类算法相结合,采用群体搜索策略,将待分类的数据对象视为抗原(Ag),把聚类中心看作抗体(Ab),通过免疫系统不断产生抗体,识别抗原来优化FKCM的目标函数,能快速地获得全局最优解。仿真结果证明了该算法在变压器故障诊断上的可行性和有效性。Considering the fuzzy clustering and kernel clustering algorithm applied in DGA analysis of power transformer existing problems of sensitive to initial value and easy to fall into local optimum,a new algorithm of fuzzy kernel clustering based on artificial immune is proposed.The algorithm combines clone selection principle based immune cloning algorithm with mature affinity and fuzzy kernel clustering,which applied group search strategy.The algorithm treats classification data objects and clustering center as antigen （Ag） and antibodies （Ab）,respectively,and it optimizes FKCM target function through producing antibodies constantly by immune system and identifying antigen,which can achieve global optimal solution quickly.Simulation results verified the feasibility and effectivity of the proposed algorithm applied in the DGA analysis of power transformer.

关 键 词：电力变压器 溶解气体分析法 模糊核聚类 人工免疫 群体搜索 

分 类 号：TM401.1[电气工程—电器]