小波包系数能量百分比在局部放电模式识别中的应用研究  

Application of Wavelet Packet Coefficients Energy Percentage in Pattern Recognition of Partial Discharge

作　　者：吴炬卓 牛海清[2] 许佳[2] 

机构地区：[1]珠海供电局,广东珠海519000 [2]华南理工大学电力学院,广东广州510640

出　　处：《电器与能效管理技术》2017年第8期12-16,28,共6页Electrical & Energy Management Technology

基　　金：国家863计划资助项目(2011AA05A120)

摘　　要：在对试验获得的4种放电信号波形进行去噪基础上,进行小波包分解并在各频带上计算小波包系数能量百分比,将其作为特征向量输入支持向量机进行放电识别。作为比较,同时将特征向量输入到BP神经网络进行识别。识别结果表明,小波包系数能量百分比构成的特征向量能够很好地反映原始信号的特征,且基于支持向量机较基于BP神经网络具有更好的识别效果。Four kinds of discharge waveforms are denoising processed, based on it , wavelet packet wasdecomposed the percentage of energy of wavelet packet coefficients was caleulated in different frequency bands, which as the feature vectors,was input into SVM and BP neural network to be recognized. The recognition results show that the eigenvectors composed by the percentage of energy of wavelet packet coefficients can well reflect thecharacteristics of the original signal,and that based on SVM has better recognition performance than the BP neuralnetwork.

关 键 词：局部放电 小波包分解 能量百分比 特征提取 支持向量机 模式识别 

分 类 号：TM20[一般工业技术—材料科学与工程]