摘要
局部放电(简称局放)测试是开关柜绝缘缺陷检测的有效手段之一。通过运用超声波局放信号的特征提取方法实现对绝缘缺陷类型的高效识别。首先运用自适应(LMS)-小波软阈值法对信号作滤噪处理,然后提取时域特征、小波包变换系数能量熵以及小波包分解系数特征,并采用最大相关最小冗余原则选取特征,由选取特征建立基于概率神经网络(PNN)的训练样本,最后导入测试样本确认识别准确度。测试结果表明:通过本方法可以高效地识别绝缘缺陷类型,其识别准确率达90%以上,对基于超声波局放信号的绝缘缺陷类型识别有一定参考价值。
Partial discharge(pd) test is one of the effective methods to detect insulation defects in switchgear. This paper uses the feature extraction method of ultrasonic local emission signal to realize the efficient identification of insulation defect types. First using least mean square method(LMS)-Wavelet soft threshold method filter to deal with the noise signal, and then extracting the time domain characteristics, the energy entropy of wavelet packet transform coefficient and the characteristics of wavelet packet decomposition coefficients, and uses the principle of maximum correlation minimum redundancy selection characteristics, set up by the selected features based on probabilistic neural network(PNN) of training samples, the final import test sample to confirm the identification accuracy. The test results show that the method in this paper can effectively identify the insulation defect types, and its recognition accuracy is more than 90%.
作者
郑祥
管鹏
田伟
Zheng Xiang;Guan Peng;Tian Wei(School of Electrical and Informational Engineering,Dalian Jiaotong University,Dalian 1160028,China)
出处
《电子测量技术》
2020年第16期124-127,共4页
Electronic Measurement Technology