以复小波变换系数为特征量的局放模式识别

复小波变换在时域和频域都具有表征信号局部特征的能力,且复小波变换提供了独有的相位信息.文中对4种典型绝缘缺陷产生的局部放电脉冲波形进行复小波变换,用模糊聚类的方法分别对各尺度复小波系数的实部R、虚部I以及复合信息系数R|I|聚类分析,将聚类的能量作为模式识别的特征量.通过大量的实验获得放电样本,用构建的BP神经网络作为分类器,对4种典型绝缘缺陷产生的局部放电进行了有效识别,结果表明:从复小波的复合信息系数R|I|中提取的特征量优于从实部R、虚部I以及实小波系数中提取的特征量.

PD Pattern Recognition Using Complex Wavelet Transform Coefficient As The Feature

Complex wavelet transform can characterize the partial feature of the PD signal in time-domain and frequency-domain,and provides the unique phasic information.In this paper,the PD pulse waveforms which are created by 4 typical insulated defects are transformed by complex wavelet,and then the complex wavelet coefficient's real part,imaginary part and compound coefficient are clustered by the Fuzzy c-means,the energy of the cluster is the feature of pattern recognition.Discharge samples are got through large number of experiments,and BPNN can identify the PD created by 4 typical insulated defects effectively.The results show that the feature extracted from compound coefficient is better than the feature extracted form the real part and imaginary part of complex wavelet coefficient or wavelet coefficient.