基于GJO特征量优选的AO-RF的变压器故障诊断模型  

Fault Diagnosis Model of Transformer Based on GJO Feature Optimization and AO-RF

作　　者：叶育林 刘森 黄松 韩晓慧 杜振斌 李彬[4] 吕杰 薛杨 赵春琳 YE Yulin;LIU Sen;HUANG Song;HAN Xiaohui;DU Zhenbin;LI Bin;LYU Jie;XUE Yang;ZHAO Chunlin(China Nuclear Power Engineering Co.,Ltd.,Guangzhou Shenzhen 518124,China;School of-Electrical-Engineering,Hebei University of-Science and-Technology,Shijiazhuang 050018,China;Hebei Provincial Key Laboratory of-Electromagnetic&Structural Performance of-Power Transmission and-Transformation Equipment,Baoding Tianwei Baobian Electric Co.,Ltd.,Hebei Baoding-071056,China;Baoding Tianweixinyu Technology Development Co.,Ltd.,Hebei Baoding 071056,China;Hebei Weixun Electric Power Automation Equipment Co.,Ltd.,Hebei Hengshui-053000,China)

机构地区：[1]中广核工程有限公司,广东深圳518124 [2]河北科技大学电气工程学院,石家庄050018 [3]保定天威保变电气股份有限公司河北省输变电装备电磁与结构性能重点实验室,河北保定071056 [4]保定天威新域科技发展有限公司,河北保定071056 [5]河北卫讯电力自动化设备有限公司,河北衡水053000

出　　处：《高压电器》2024年第5期99-107,共9页High Voltage Apparatus

基　　金：河北省省级科技计划资助(20312101D)。

摘　　要：在变压器故障诊断过程中,进行合理的特征优选,将有助于提高诊断模型的诊断精度,为此,文中提出了一种基于金豺优化算法(golden Jackal optimization,GJO)特征量优选与AO-RF的变压器故障诊断模型。首先,采用GJO对构建的21维变压器油中溶解气体特征量进行优选;然后,根据GJO得到的特征优选结果,采用天鹰算法(aquila optimizer,AO)优化随机森林(random forest,RF)的变压器故障诊断模型对变压器故障进行诊断,并与不同特征量、不同故障诊断模型的诊断结果进行了对比。实验结果表明:GJO优选特征量相比21维原始特征、三比值法、无编码比值法以及AO优选特征量的故障诊断准确率可提高1.12%~25.78%,kappa系数可提高0.02~0.24;AO-RF故障诊断模型较RF、SVM、ELM、SSA-RF、WOA-RF、GJO-RF模型的诊断准确率可提高1.84%~15.86%,kappa系数可提高0.02~0.16,验证了所提方法的有效性和准确性。In the fault diagnosis process of transformer,reasonable feature selection is contribute to improve the diagnostic accuracy of the diagnosis model.For that,a fault diagnosis model of transformer based on GJO feature optimization and AO-RF is proposed in this paper.Firstly,the golden Jackal optimization(GJO)algorithm is used to optimize the characteristic quantity of 21-d transformer dissolved gas-in-oil;Then,according to the feature optimization results obtained by GJO,the fault diagnosis model of transformer with random forest(RF)optimized by the aquila optimization(AO)algorithm is used to diagnose the fault of transformer.And it is compared with the diagnosis results that have different feature parameters and different fault diagnosis models.The experimental result shows that the fault diagnosis accuracy of the GJO optimized feature,compared with 21-d original feature,three-ratio method,non-coded ratio method and AO optimized feature,can be improved by 1.12%-25.78%,and the Kappa coefficient can be increased by 0.02-0.24;The fault diagnosis accuracy of AO-RF,compared with RF,SVM,ELM,SSA-RF,WOA-RF and GJO-RF models,can be increased by 1.84%-15.86%,and the Kappa coefficient can be increased by 0.02-0.16,which verify the effectiveness and accuracy of the proposed method.

关 键 词：变压器 故障诊断 金豺算法 随机森林 天鹰算法 

分 类 号：TM407[电气工程—电器] TP18[自动化与计算机技术—控制理论与控制工程]