摘要
提出了一种基于K近邻KNN(K-Nearest Neighbour)算法的换流变压器故障诊断方法。设计了4种人工油纸绝缘缺陷,采用超高频天线采集局部放电信号。通过对局部放电超高频信号进行小波包多尺度变换,计算其多尺度小波系数的能量系数。采用KNN算法对局部放电超高频信号能量特征参数进行识别。将反向传播神经网络和所提方法对局部放电超高频信号模式的识别结果进行了对比,结果表明所提出的方法更适用于换流变压器故障诊断。
A kind of fault diagnosis method based on KNN （K-Nearest Neighbour） is proposed for converter transformer. Four types of artificial oilpaper insulation defect are designed and the UHF（Ultra High Frequency） antenna is used to collect PD（Partial Discharge） UHF signals. Multi-scale wavelet packet transform is carried out for calculating the energy coefficients of collected PD UHF signals and the KNN algorithm is applied to recognize their characteristic parameters. The PD mode recognized by the BPNN（Back Propagation Neural Network ） is compared with that by the proposed algorithm,which shows that the latter is more suitable for the fault diagnosis of converter transformer.
出处
《电力自动化设备》
EI
CSCD
北大核心
2013年第5期89-93,共5页
Electric Power Automation Equipment
关键词
换流变压器
局部放电
超高频
故障诊断
故障分析
小波分解
converter transformer
partial discharges
ultra high frequency
fault diagnosis
failure analysis
wavelet decomposition
