基于快速独立分量分析算法的气体绝缘开关设备局部放电混合信号分离与缺陷类型辨识  

Separation of Partial Discharge Mixing Signals and Type Identification of Defects in Gas Insulated Switchgear Based on Fast Independent Component Analysis Algorithm

作　　者：云玉新 赵笑笑 李世鹏[3] 李可军[3] 张晓星[4] 

机构地区：[1]国网山东省电力公司电力科学研究院,济南250002 [2]国网技术学院,济南250002 [3]山东大学电气工程学院,济南250061 [4]重庆大学输配电装备及系统安全与新技术国家重点实验室,重庆400030

出　　处：《高电压技术》2014年第3期853-860,共8页High Voltage Engineering

基　　金：国家重点基础研究发展计划(973项目)(2009CB724506)~~

摘　　要：为解决气体绝缘开关设备(GIS)内多种绝缘缺陷并存时的混合局部放电(简称局放)的信号分离和缺陷类型辨识问题,提出了基于快速独立分量分析(FastICA)算法的GIS局放混合信号分离与模式识别方法。该方法具体包括:采用FastICA算法从局放混合信号中分离出单一局放信号,针对各单一局放信号,提出的模式识别策略是通过增添相反符号和幅值归一化的单一局放信号用于分类器训练或者使用对信号符号不敏感的特征(如分形特征等)作为分类特征。算例分析结果表明,提出的方法能够实现GIS多绝缘缺陷局放信号的有效分离,并能对分离出的单一局放信号进行准确缺陷类型辨识,且具有抗噪能力强、计算速度快、robust性好等优点。In order to realize the signals separation and pattern recognition of mixing partial discharge（PD） that exist with multiple insulation defects in gas-insulated metal-enclosed switchgear（GIS）, we presented a method based on fast inde- pendent component analysis （FastlCA） algorithm. In detail, we separated single PD signals from mixing signals using the FastlCA algorithm, then proposed a recognition strategy that additional pole-reversed and amplitude-normalized single insulation defect signals were used to train classifiers or the characteristics insensitive to signal polarity were used as clas- sification characteristics. The calculation and analysis of an sample show that, the proposed method can separate PD mixing signals effectively and then recognize single insulation defects based on the separated single PD signals, and is insensitive to noise, fast in calculation and good in robustness.

关 键 词：气体绝缘开关设备 局部放电 快速独立分量分析 多绝缘缺陷 混合信号分离 缺陷类型辨识 

分 类 号：TM564[电气工程—电器]