DGA与GRNN的联合变压器故障诊断研究  

Study of transformer fault diagnosis based on GRNN and DGA

作　　者：丁硕[1] 常晓恒[1] 巫庆辉[1] 杨友林[1] 

机构地区：[1]渤海大学工学院,锦州121013

出　　处：《电子测量技术》2014年第5期142-146,共5页Electronic Measurement Technology

基　　金：国家自然科学基金资助项目(61104071)

摘　　要：针对传统的变压器故障诊断方法的不足,提出了基于油中溶解气体分析(DGA)方法和广义回归神经网络(GRNN)的变压器故障诊断方法。以DGA方法获取GRNN故障诊断模型的输入特征向量,建立了GRNN故障诊断模型,为了检验GRNN诊断模型的实际诊断能力,以某变电所主变压器的4种典型故障诊断为例进行仿真实验,并与标准BP神经网络(BPNN)和LM算法改进的BPNN(LM-BPNN)的诊断结果进行对比。仿真结果表明:DGA方法与GRNN的联合变压器故障诊断方法的诊断速度更快、准确率更高和泛化能力更强,且GRNN故障诊断模型构建简单,验证了所提出方法的实用性和有效性。In view of disadvantages of traditional transformer fault diagnosis techniques,fault diagnosis methods based on DGA and GRNN are proposed.Input characteristics vectors of GRNN fault diagnosis model are obtained by DGA and GRNN fault diagnosis model is established.To test the model＇s practical diagnosis ability,diagnoses of four typical transformer faults in a certain substation are taken as examples in a simulation experiment.The diagnosis results are compared with the results of standard BPNN and LM-improved BPNN.The simulation results show that the transformer fault diagnosis method based on GRNN and DGA has faster diagnosis speed,higher classification accuracy,stronger generalization ability and the establishment of the model is simple,so the practicality and effectiveness of the method proposed is verified.

关 键 词：DGA方法 GRNN 标准BPNN LM-BPNN 故障诊断 泛化能力 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TM411[自动化与计算机技术—控制科学与工程]