摘要
为了检验基于小波包-反向传播神经网络的局部放电信号模式识别方法的有效性,设计了4种典型的绝缘缺陷模型,并采用局部放电传感器测量了局部放电信号。采用小波包分解提取了局部放电信号在多尺度上的小波包系数,将计算得到的多尺度小波包系数的能量参数作为特征参数,结合反向传播神经网络对不同类型的放电进行模式识别。研究结果表明,基于小波包-反向传播神经网络的模式识别方法对变压器油纸绝缘局部放电有较强的识别能力,且5层小波包分解层数的识别率高于4层小波包分解层数的识别率。
In order to verify the validity of partial discharge signal pattern recognition method based on wavelet packet-back propagation neural network, four typical transformer insulation defect models were designed and partial discharge sensor was used to measure the partial discharge signals. Wavelet packet decomposition was used to extract the wavelet packet coefficients of partial discharge signals on multiple scales, and the energy parameters of the calculated multi-scale wavelet packet coefficients were taken as characteristic parameters. Combined with back propagation neural network, different types of discharge were identified. The results show that the method based on wavelet packet and back propagation neural network has a strong ability to recognize the partial discharge of transformer oil-paper, and the recognition rate of 5 decomposition layers of wavelet packet is higher than that of 4 decomposition layers of wavelet packet.
作者
于建军
YU Jianjun(Yunfeng Power Plant,Ji’an 134200,China)
出处
《微型电脑应用》
2021年第3期128-130,共3页
Microcomputer Applications
关键词
油纸绝缘
局部放电信号
小波包变换
反向传播神经网络
模式识别
oil-paper insulation
partial discharge signal
wavelet packet transform
back propagation neural network
pattern recognition