
2016年第44卷第11期 2016,Vol.44No.11
陕西电力
SHAANXI ELECTRIC POWER
特别推荐 Featured Focus
基于振动-SVM的变压器绕组缺陷诊断方法
张琳",马宏忠",王涛云2,李勇3,许洪华3
(1.河海大学能源与电气学院,江苏南京211100;2.上海市电力公司全山供电公司,上海201500:
3.江苏省电力公司南京供电公司,江苏南京210008
摘要;为了诊断变压器绕组是否存在松动缺陷以及判断绕组松动严重程度,给出及时应对猎施,提出以频率特征量为输入,PSO优化SVM的变压器绕组松动缺陷诊断方法。进行110kV变压器短路试验,模拟不同程度的绕组松动缺陷以探究绕组松动程度特征,提取振动信号的频谱分量,利用比值法削弱负载电流的影响,发现基频分量、800Hz和900Hz分量与频率分量总和的比值随绕组松动程度增加而产生明显变化。但由于特征值相对较小,不利于提取判断阔值,故通过支持向量机进行缺陷诊断并利用效果相对较好的粒子群算法对其进行优化。测试结果表明该诊断模型具有较高的准确性。
关键词:变压器;绕组松动;频谱分析;支持向量机
中图分类号:TM835
文献标志码:A
文章编号:1673-7598(2016)11-0014-05
DefectDiagnosisforTransformerWindingBased onVibration-SVM
ZHANG lin', MA Hongzhong', WANG Taoyun?, LI Yong", XU Honghua(1.College of Energy and Electrical Engineering, Hohai University, Nanjing 211100, China;
2.StateGrid Jinshan PowerSupply Company,Shanghai201500,China; 3.StateGrid Nanjing Power Supply CompanyNanjing210008,China)
Abstract: In order to estimate the loosing deformation in transformer winding and determine the degree of the loosing deformation, the paper proposes a transformer fault diagnosis method using frequency spectrum characteristic as input and particle swarm optimization (PSO) to optimize the support vector machine(SVM). At first, this paper conducts 110 kV transformer shortcircuit test,simulates the different degree of winding looseness to explore its featureextracts the spectral component of vibration signal, reduces the effects of load current by ratio method, and finds out that the ratio of 100 Hz, 800 Hz and 900 Hz to the sum of frequency components changes significantly with increasing in the degree of the winding looseness.But considering that these characteristic values are too small to extract threshold values,the SVM is used to diagnose the defect and the PSO is used to optimize
the SVM. The result shows that PSOSVM model can enhance the aceuracy in judging winding loosing Key words: transformer; winding looseness; frequency spectrum analysis; support vector machine
由于绕组松动存在累积效应,如果绕组在短路
引言 o
变压器是电力系统中最重要的设备之一,是电力传输的枢纽,其运行的安全与否将直接决定整个电网是否能够安全运转。变压器出厂后,其运输、安装、短路电流冲击等都可能导致其绕组压紧力减小,造成绕组松动,长期累积会严重威变压器的正常运行。
基金项目:国家自然科学基金(51177039)万方数据
电动力作用或其他原因作用下发生轻微松动,其机械性能和绝缘性能都会有所下降,在下次短路电流的冲击下,绕组松动程度会再次增加,其性能也会继续恶化,陷人恶性循环3,所以发现变压器绕组松动及准确判断绕组松动程度有助于排除变压器存在的隐患,采取正确的应对措施。文献[4]中提出绕组发生松动时,基频分量幅值会发生改变,但只判断了绕组是否发生松动,未对绕组松动程度进行研究。文献[5]提出用小波神经网络诊断绕组和铁心故障,但是需要大量训练样本提高判断精度。目前其它常用