基于放电时差的局部放电模式识别的研究  

PD Pattern Recognition Using Inter-time

作　　者：刘玲[1] 廖瑞金[1] 周湶[1] 叶开颜[1] 李剑[1] 

机构地区：[1]重庆大学高电压与电工新技术教育部重点实验室,重庆400044

出　　处：《高电压技术》2007年第8期35-39,共5页High Voltage Engineering

基　　金：国家杰出青年基金(50425722);重庆市自然科学基金(11699)。~~

摘　　要：为寻找一种无需校正、方便信号监测的局部放电模式识别方法,将局部放电脉冲间的时间差分布引入到局放放电的模式识别中,构造了放电相位、时间差与放电次数分布的三维谱图Hn(Δt,)φ,并分析提取了其灰度图象的盒维数与信息维数特征参量,最后以分形维数作为输入,径向基函数神经网络(RBFNN)作为模式分类器对5种人工油纸绝缘缺陷模型进行识别。研究表明,识别率均>90%并具有较强的抗干扰能力。Auto-recognition to discharge types in on-line partial discharge （PD） monitoring system could be used to find out internal partial defects and the relevant discharge development degree in time, and then prevents equipment from the coming faults. Time intervals between consecutive discharges （△t） may even be more efficient to characterize a defect or to differentiate between different defects. In this paper, five kinds of typical artificial defect models of oil-paper insulation, such as oil-gap discharge model, surface discharge model, cavity discharge model, corona discharge model and mixture discharge model were designed, PD sample data caused by artificial defects were got by PD detection system in laboratory. Utilizing these sample data, the inter-time distributions of PD pulse were introduced in PD pattern recognition, the 3-D image Hn （△t, Ф） of discharge phase, inter-time and discharge number Ф-△t- n distribution was constructed, and the box dimension and information dimension of gray intensity images which were transferred by 3-D images were analyzed and extracted. In this scheme, PD fractal dimensions were used as in- puts, radial basis function neural network （RBFNN） which was prior to BP Neural Network in the ability of approach, the ability of classification and the rate of train was used as classifier, five kinds of artificial oil-paper insulation defects were distinguished. The results show recognition rates are all over 90% with a strong noise interference suppression.

关 键 词：局部放电 放电时差 放电相位 模式识别 油纸绝缘 分形维数 RBF神经网络 

分 类 号：TM835[电气工程—高电压与绝缘技术]