GIS典型缺陷的局部放电超高频检测及模式识别  

UHF Detection and Pattern Recognition of Partial Discharge on GIS Typical Defects

作　　者：韩磊[1] 王立威[2] 郑艳清 

机构地区：[1]内蒙古电力科学研究院,呼和浩特010020 [2]西安电子科技大学机电学院,西安710126 [3]内蒙古国合电力有限责任公司,呼和浩特010020

出　　处：《内蒙古电力技术》2015年第1期7-12,共6页Inner Mongolia Electric Power

基　　金：内蒙古电力(集团)有限责任公司2012年第二批科技项目(51014112003)

摘　　要：为了解决GIS局部放电带电检测难题,采用了超高频法,以GIS试验设备为对象,设计和模拟了GIS中自由金属颗粒、悬浮电位体、母线金属尖端、外壳金属尖端、绝缘子表面金属颗粒和绝缘子气泡6种典型缺陷的模型,基于超高频法对其放电信号进行检测,提取了缺陷特征参数,应用支持向量机进行模式识别,进而对支持向量机的惩罚参数"C"和核函数参数"g"进行了粒子群优化。试验结果表明:不同缺陷类型的超高频信号在图谱和提取的数据中会呈现出不同的特征;在模式识别中,粒子群方法优化支持向量机展现出了比支持向量机更好的鲁棒性和泛化能力。The ultra-high frequency（UHF） method was applied for partial discharge detection in GIS, taking experimental GIS equipment in the laboratory as experimental subject, the typical defects including free particles, metal spikes, floating potential, insulators defect were designed and simulated in the GIS. The UHF method was used to detect its discharge signal, and extracted the parameters of the defect characteristics. The support vector machine was used for pattern recognition, and particle swarm optimization was carried out for support vector machine penalty parameter“C”and kernel function parameter“g”. The results showed that the UHF signal of different types of defects in the spectrum and the extracted data would show different characteristics in the UHF detection;particle swarm optimization support vector machine parameters showed better robustness and generalization ability than the support vector machine in pattern recognition.

关 键 词：气体绝缘组合电器 局部放电 超高频 支持向量机 粒子群优化 

分 类 号：TM621.8[电气工程—电力系统及自动化]