基于粒子群K均值算法的变压器在线故障诊断研究  

The research on the on-line fault diagnosis of transformer based on PSO and K means algorithm

作　　者：鄢小虎[1,2] 何发智[2] 卢文华[1] 谷凯凯[1] 向冬冬[1] 文正其[1] 

机构地区：[1]国网电力科学研究院武汉南瑞有限责任公司,武汉430074 [2]武汉大学计算机学院,武汉430072

出　　处：《电测与仪表》2016年第4期1-4,共4页Electrical Measurement & Instrumentation

基　　金：国家973计划资助项目(2011CB707904);国家自然科学基金资助项目(61070078);国家电网公司科技项目(521997140005)

摘　　要：文章提出了基于粒子群K均值算法的变压器在线故障诊断方法。首先通过K均值算法得到代表各种变压器故障类型的三比值,然后利用粒子群算法进行优化。当改良的三比值法由于缺码不能进行分析时,计算待判样本的三比值到各类故障对应三比值的距离,选择距离最小的三比值对应的故障类型为该样本的故障。然后结合在线监测数据,利用专家规则库融合多参量对故障进行综合诊断。最后通过实例验证了文章方法的可行性和有效性。The on-line fault diagnosis of transformer based on PSO and K means algorithm is proposed in this paper. Firstly, the three rations data of oil chromatogram is clustered by the K means algorithm. Then, the three rations that can represent every type of the transformer fault are optimized using PSO. When the diagnosis cannot be analyzed by the improved three-ration method as code deficiency, the distances between the three rations of the oil chromatographic sample will be analyzed and the optimal three rations of every fault type are computed. The fault type with the nearest distance is the final fault type of the sample. Secondly, combined with the multi-parameter on-line monitoring data, the transformers fault is comprehensively diagnosed with experts rule database. Finally, the feasibility and efficiency of the method proposed in this paper is demonstrated by the experiment.

关 键 词：粒子群算法 K均值算法 油色谱 故障诊断 改良三比值法 

分 类 号：TM411[电气工程—电器]