基于波形非线性映射的多局部放电脉冲群快速分类  

The Fast Grouping Technique of PD Sequence Based on the Nonlinear Mapping of Pulse Shapes

作　　者：司文荣[1] 李军浩[1] 袁鹏[1] 李延沐[1] 梁永春[1] 李彦明[1] 

机构地区：[1]西安交通大学电气工程学院,西安710049

出　　处：《电工技术学报》2009年第3期216-221,228,共7页Transactions of China Electrotechnical Society

基　　金：河北省自然科学基金资助项目(F2007000636)

摘　　要：简单分析了目前基于脉冲峰值-时间序列的局部放电检测系统在模式识别功能上存在的缺陷。提出利用基于单个脉冲波形的宽带检测技术研制多局放检测与模式识别系统,并给出了研制交流下该系统的数据处理方案。重点对研制该系统需要解决的脉冲群快速分类技术进行了阐述:即通过基于局放脉冲时域波形的非线性映射进行参数提取,形成2D或3D的波形特征参数空间,再使用非监督的模糊聚类分析实现局放脉冲群的快速分类。基于2D特征参数平面的仿真试验和3D特征参数空间的GIS污秽试验数据处理结果均表明该分类技术的可行性和实用性。这为多局放电检测与模式识别系统的研制提供了试验和理论依据。To overcome the disadvantage of partial discharge（PD） pattern recognition method based on pulse peak-time sequence, a novel multi-PDs detection and pattern recognition system developed by wideband detection technology based on PD pulse waveshape is proposed. The data processing program of the system development under AC voltage is given. Focus is made on the fast pulse sequence grouping technique based on PD pulse waveshape. It is realized with feature extraction by the nonlinear mapping of PD pulse waveshapes in time domain, making the 2D parameters plane or 3D parameters space, then using the unsupervised learning fuzzy clustering to achieve fast grouping for PD pulse sequence. If separation is performed successfully, the original PD pulse sequence can be split in more sub-sequences, each one relevant to a well defined PD phenomenon or electrical noise. Test results show that the grouping technique is feasible and practical, which provides a good method to develop the multi-PDs detection and pattern recognition system.

关 键 词：多局放 脉冲波形 非线性映射 特征提取 模糊聚类 快速分类 

分 类 号：TM835[电气工程—高电压与绝缘技术]