基于机器学习与DGA的变压器故障诊断及定位研究  

Study of Transformer Fault Diagnosis and Location Based on Machine Learning and DGA

作　　者：周光宇 马松龄[1] ZHOU Guangyu;MA Songling(School of Mechanical and Electrical Engineering,Xi’an University of Architecture and Technology,Xi’an 710055,China)

机构地区：[1]西安建筑科技大学机电工程学院,西安710055

出　　处：《高压电器》2020年第6期262-268,共7页High Voltage Apparatus

摘　　要：变压器故障诊断特征信息繁多,且故障点难以确定,为有效利用故障信息提高故障诊断准确率,以及实现故障定位,提出一种基于粗糙集知识和优化支持向量机的变压器分层故障诊断及定位新方法。首先使用邻域粗糙集评估DGA样本重要度,并约简出优选故障诊断特征量。其次构建基于多分类支持向量机的分层故障诊断模型,采用粒子群算法优化模型参数以提高分类精度,实现了故障性质和故障定位的多层诊断。实例分析表明,新特征量可以提高机器学习的知识挖掘能力,不仅故障分类的精度增加,而且模型可以实现故障点的定位,综合诊断准确率达到88.4%。There are many fault diagnosis features of the transformer,and difficult to determine the point of failure.In order to effectively use fault information,improve the accuracy of fault diagnosis and achieve fault location,con⁃structed a new method for fault diagnosis stratified and location of transformers based on Rough set knowledge and optimized Support Vector Machine.First,used the neighborhood rough set to assess DGA sample importance and re⁃duced to get optimal feature quantity.Secondly,built a hierarchical fault diagnosis model based on multi⁃class sup⁃port vector machine,used the particle swarm optimization parameter selection is used to improve the accuracy,achieved stratified multi⁃layered nature of the fault diagnosis and fault location.The example analysis shows that new feature quantity can improve the knowledge mining ability of machine learning,not only increases the accuracy of fault classification,and the model can realize the positioning of the fault point,comprehensive diagnostic accuracy reached 88.4%.

关 键 词：变压器 故障诊断 邻域粗糙集 支持向量机 多层诊断 故障定位 

分 类 号：TM407[电气工程—电器]