摘要
根据气体绝缘组合电器(GIS)设备内部绝缘缺陷产生局部放电的特点,设计了4种典型的GIS缺陷模型,采用甚高频高速采集大量局部放电样本,构造了局部放电图谱;以差盒维数和多重分形理论为基础,给出了基于差盒维数的多重分形计算方法;提出了一种基于多重分形特征的GIS局部放电图谱特征提取方法,对局放图像求取了相应的差盒维数、多重分形维数及放电重心特征,最后将提取的特征量通过RBF神经网络进行分类,识别结果显示本文方法有效地提高了GIS局部放电4种缺陷的识别率。
Aiming at the internal isolation defects in were designed. The GIS gray intensity images were GIS and PD characteristics, four kinds of GIS defection models constructed based on mass discharge specimens gathered by the ultra-high frequency and high speeds systems. The multi-fractal dimensions were founded based on the box-counting dimension and muhi-fractal theories. The GIS gray intensity image extraction method based on the muhi-fractal characteristics is putted forward. The box-counting dimension,muhi-fractal dimensions and discharge centrobaric characteristics of the PD pictures are also extracted, and the characteristic variables are classified by RBF network. The identification results show that the proposed method can effectively improve the discrimination rates for four kinds of defects in PD
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2007年第4期597-602,共6页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金(50577069)
重庆市自然科学基金(CSTC2005BB3170)资助项目
