深度学习神经网络在电力变压器故障诊断中的应用  

Application of Deep Learning Neural Network in Fault Diagnosis of Power Transformer

作　　者：石鑫[1] 朱永利[1] 

机构地区：[1]华北电力大学控制与计算机工程学院,河北省保定市071003

出　　处：《电力建设》2015年第12期116-122,共7页Electric Power Construction

基　　金：河北省自然科学基金项目(E2009001392)

摘　　要：由于电力变压器发生故障时油色谱在线监测数据无标签,工程现场往往会得到大量无标签故障样本,而传统的故障诊断方法在对变压器故障类型进行判别时往往无法充分利用这些无标签故障样本。该文基于深度学习神经网络(deep learning neural network,DLNN),构建了相应的分类模型,分析并用典型数据集对其分类性能进行测试。在此基础上提出一种电力变压器故障诊断新方法,它能够有效利用大量电力变压器油色谱在线监测无标签数据和少量故障电力变压器油中溶解气体分析(dissolved gas-in-oil analysis,DGA)实验数据进行训练,并以概率形式给出故障诊断结果,具有更优的故障判别性能,能够为变压器的检修提供更为准确的参考信息。工程实例测试结果表明,该方法正确可行,诊断性能优于三比值、BP神经网络和支持向量机的方法,适用于电力变压器的故障诊断。As oil chromatography online-monitoring data is unlabeled during pow er transformer failure,project sites tend to get a large number of unlabeled fault samples. How ever,traditional diagnosis methods often fail to make full use of those unlabeled fault samples in judging transformer fault types. Based on deep learning neural netw ork（ DLNN）, a corresponding classification model w as established,w hose classification performance w as analyzed and tested by typical datasets. On this basis,a new fault diagnosis method of pow er transformer w as further proposed,in w hich a large number of unlabeled data from oil chromatogram on-line monitoring devices and a small number of labeled data from dissolved gas-inoil analysis（ DGA） w ere fully used in training process. It could generate fault diagnosis result in the form of probabilities,and provide more accurate information for the maintenance of pow er transformer because of its better performance in fault diagnosis. Testing results from engineering example indicate that the proposed method is correct and feasible,and its diagnosis performance is better than that of three radio,BP neural netw ork and support vector machine,w hich is suitable for the fault diagnosis of pow er transformer.

关 键 词：电力变压器 故障诊断 油中溶解气体分析 深度学习神经网络 

分 类 号：TM721[电气工程—电力系统及自动化]