采用小波包树能量矩阵奇异值分解的局部放电模式识别

发布时间：2018-06-23 14:31

本文选题：局部放电识别 + 能量特征提取　；参考：《西安交通大学学报》2017年08期

【摘要】：为了从局部放电信号电磁波大量冗余的频率信息中提取有效的局部放电特征参量,提出了采用小波包和奇异值分解相结合的能量特征提取方法。该方法从大量小波包树节点的能量信息中提取了一组能量特征参量用于局部放电类型识别,首先对4种放电类型的电磁波信号进行小波包分解,计算每个小波包节点系数的能量,接着采用奇异值分解法从小波包树所有节点的能量信息中提取奇异值较大的一组参量,利用这组参量建立识别模型,最后使用支持向量机对4种变压器典型放电类型进行识别。结果表明,小波包树能量矩阵奇异值分解可以从包含有大量无效和冗余频率信息的电磁波信号中提取能量特征参量,从而进行局部放电识别。研究内容可为局部放电类型识别提供一种有效的特征提取方法。
[Abstract]:In order to extract the effective characteristic parameters of partial discharge from the redundant frequency information of electromagnetic wave of partial discharge signal, an energy feature extraction method combining wavelet packet and singular value decomposition is proposed. In this method, a set of energy characteristic parameters are extracted from a large number of energy information of wavelet packet tree nodes for partial discharge type identification. Firstly, four types of electromagnetic wave signals are decomposed by wavelet packet. The energy of each wavelet packet node is calculated. Then the singular value decomposition method is used to extract a group of parameters with large singular value from the energy information of all nodes in the wavelet packet tree. Finally, support vector machine is used to identify four typical discharge types of transformers. The results show that the singular value decomposition of the energy matrix of wavelet packet tree can extract the energy characteristic parameters from the electromagnetic wave signal which contains a large number of invalid and redundant frequency information, so as to identify the partial discharge. The research can provide an effective feature extraction method for partial discharge type recognition.
【作者单位】：西安热工研究院有限公司;中国华能集团公司;西安交通大学电力设备电气绝缘国家重点实验室;
【基金】：国家自然科学基金资助项目(51677143,51607138)
【分类号】：TM855

【参考文献】