统计参数在变压器局部放电模式识别中的应用  

Application of Statistic Parameters in Recognition of Partial Discharge in Transformers

作　　者：胡文堂[1] 高胜友[2] 余绍峰[1] 谈克雄[2] 高文胜[2] 

机构地区：[1]浙江省电力试验研究院,杭州310014 [2]清华大学电机系电力系统及发电设备控制和仿真国家重点实验室,北京100084

出　　处：《高电压技术》2009年第2期277-281,共5页High Voltage Engineering

基　　金：基金资助项目:国家自然科学基金(59977011)~~

摘　　要：对局部放电进行模式识别可以了解放电类型和严重程度,为故障诊断和检修提供参考依据。为此将Weibull统计分布参数用于局部放电模式识别当中,采用仿真分析和模型试验的方法证明了局部放电的脉冲高度分布符合Weibull统计分布规律。在统计放电脉冲高度分布时对放电幅值进行了归一化处理,将放电累积概率为99%的放电幅值作为归一化因子,以消除试验中偶然出现的大的随机放电脉冲干扰的影响。使用工频电压正负半周的Weibull统计分布的形状参数、放电幅值中心和放电相位中心共6个参数作为特征向量,以人工神经网络为分类器,对放电类型获得了超过85%的识别率。研究表明,这种故障模式的表征方法具有模式特征数量少、表征能力强等优点,采用人工神经网络方法可以准确识别不同模式的放电,具有较高的识别率。The pattern recognition for partial discharge （PD） is helpful for estimating PD type and level, which will provide a scientific basis for failure diagnosis and maintenance. We used Weibull statistic distribution parameters in the pattern recognition for PD, and adopted the simulation and model test to prove that the PD height distribution （PDHD） fits well with 2-parameter Weibull distribution. The simulation data were generated by the Monte-Carlo method and 8 typical PD models were designed for test. In the period of each test, the pulse amplitude and phase information of each PD pulse in 10 seconds were recorded. Meanwhile, the discharge amplitude was normalized during calculation of the PDHD. In order to eliminate the influence of random pulse interference, the amplitude whose cumulative probability is 99% was used as the normalization factor. Six characteristic parameters, including the shape parameter of Weibull distribution, the center of discharge amplitude and the center of discharge phase in positive half period and negative half period of power cycle, were adopted to characterize fault pattern. The artificial neuron network was used as classifier to get at least 85% recognition rate for PD types. The results show that excellent characterization Capability is obtained using a few of characteristic parameters, and different types of PD can be accurately identified using the artificial neuron network.

关 键 词：局部放电 模式识别 WEIBULL分布 统计参数 归一化 人工神经网络 

分 类 号：TM835.4[电气工程—高电压与绝缘技术]