以小波频带能量为特征量的故障电弧弧声识别方法  

Faults Arc Sound Signals Recognition Method Using Wavelet Frequency Bands Energy as Discriminating Features in Switch Cabinet

作　　者：蓝会立[1] 张认成[2] 李克讷[1] 冯平[1] 

机构地区：[1]广西科技大学电气与信息工程学院,广西柳州545006 [2]华侨大学机电及自动化学院,福建厦门361021

出　　处：《计算机测量与控制》2013年第2期532-534,共3页Computer Measurement &Control

基　　金：福建省高新技术研究开发计划重点项目(2005H036);福建省科技计划重大项目(2012H6013);广西高等学校科研项目(201204LX263)

摘　　要：通过故障电弧实验平台获取大量电弧声数据样本,将电弧声信号进行3层小波包分解,以小波包敏感频带能量作为识别特征量,利用模糊C-均值聚类算法对识别特征量进行模糊聚类,得到真假弧声的模糊分类矩阵和聚类中心,通过计算待测数据样本与已知特征弧声聚类中心的贴近度,实现真假弧声的识别,实验结果表明该方法对故障弧声的正确识别率在91%以上,识别效果良好;最后给出了基于早期特征弧声的故障电弧预测预警方案,改变目前故障电弧的事后被动检测,把故障电弧消除在事故发生之前,降低和避免了故障电弧产生时对开关设备造成的损失。Faults arc sound sample data caused by different arcing electrodes, space between electrodes, and discharge voltage in switch cabinet were got by fault arc detection system in laboratory. Utilizing these sample data, frequency hand energy features of arc sound signal are extracted based on three--layer wavelet packet decomposition. Faults Arc sound signal recognition scheme based on fuzzy C--mean clustering algorithm was put forward. In this scheme, some characteristic frequency bands energy were used as inputs of fuzzy C--mean clustering algorithm, the optimized classified matrix and clustering centers were obtained. By calculating the closeness rating between the new samples and the trained clustering center,the fault arc sound was identified. Good recognition result is obtained using the scheme. Finally, a forecast and early warning scheme of faults are has been put forward based on the characteristic frequency of are sound, which can change the traditional passive detection method to the active detection method , reduce the damage of switch apparatus by eliminating the arcing faults in the early time.

关 键 词：开关柜 故障电弧 弧声特征 预测预警 模糊识别 模糊聚类 

分 类 号：TM773[电气工程—电力系统及自动化] TP391[自动化与计算机技术—计算机应用技术]