隐式半马尔科夫模型下的变压器故障诊断方法  

Transformer Fault Diagnosis Method Based on Hidden Semi-Markov Model

作　　者：李刚[1] 米琛浩 郑顾平[1] 夏彦卫 Li Gang;Mi Chenhao;Zheng Guping;Xia Yanwe(Department of Computer, North China Electric Power University, Baoding071003, China;State Grid Hebei Electric Power Research Institute, Shijiazhuang050021, China)

机构地区：[1]华北电力大学计算机系,河北保定071003 [2]国网河北省电力公司电力科学研究院,石家庄050021

出　　处：《计算机测量与控制》2019年第5期30-34,共5页Computer Measurement &Control

基　　金：国家电网公司科技项目资助(5204DY170010);国家自然科学基金资助(51407076);中央高校基本科研业务费专项资金资助(2018MS075)

摘　　要：变压器运行过程中存在多种状态,能够正确划分运行状态,对变压器的维修和故障诊断有着重要的意义;首先,详细分析了马尔科夫链的衍生模型,并构造了隐式半马尔科夫模型(hidden semi-markov models,HSMM);然后,通过引入"微状态-宏状态"的对应关系,用于在HSMM中描述变压器运行过程中的状态;最后,建立了涵盖变压器历史状态信息,并包含特征提取、状态分类和故障识别过程的HSMM故障诊断流程;通过变压器DGA故障诊断的算例分析,结果表明所述方法的有效性。There are many states in the operation of the transformer,which can correctly divide the operating state,which is of great significance for the maintenance and fault diagnosis of the transformer.Firstly,the derivative models of Markov chain are analyzed in detail and Hidden Semi-Markov Models(HSMM)are constructed.And then,the state of the transformer operation process is described in HSMM by introducing the correspondence of "micro-state & macro-state".Finally,the HSMM fault diagnosis process covering the historical state of the transformer and including feature extraction,state classification and fault identification process is established.Through the analysis of the transformer DGA fault diagnosis,the results show the effectiveness of the proposed method.

关 键 词：隐马尔科夫 故障诊断 DGA 微状态-宏状态 

分 类 号：TM933[电气工程—电力电子与电力传动]