摘要
局部放电会引起变压器绝缘的老化和破坏,而变压器局部放电特性的研究能够很好反应变压器潜伏性缺陷,对其安全可靠运行具有重要意义,因而设计制作了模拟变压器沿面放电、气隙放电和电晕放电的3种缺陷模型,采用升压法进行相应的放电试验,通过分析油中溶解气体在局部放电发展过程中的变化规律,寻找出油中产生气体与不同局部放电的对应关系。引入局部放电的最大放电量相位特征谱图Hqmax(φ)和放电次数相位特征谱图Hn(φ),并提取基于谱图的统计特征参量,构建反向传播(back propagation,BP)神经网络和径向基函数(radical ba-sis function,RBF)神经网络对局部放电的放电类型进行初级识别,在此基础上,应用信息融合方法将初级识别结果融合油中产气特征以综合识别局部放电类型。实验结果表明,同一种溶解气体在不同放电模型中发展变化趋势是不一样的,根据统计特征参量和油中溶解气体变化规律,应用信息融合方法对变压器局部放电模式具有足够的识别能力。
Partial discharge characteristics is a very good response to internal insulation of latent defects,it has very important significance of safe and reliable operation for transformer.Aiming at the discharge properties of oil-paper insulation,we designed and experimentally researched 3 kinds of experimental models simulating discharges in electrical transformers.Moreover,we tested the corresponding discharge by the boost pressure method,collected the oil-gas data and partial discharge signal to analyze the variable law of the dissolved gases in oil during the development process of the partial discharge,and found the correspondence between the gas produced in oil and different discharge models.Using statistical method extracting characteristic parameters from phase spectrogram of maximum discharge capacity and discharge frequency,we constructed the BP neural network and RBF neural network to primarily recognize the discharge type of partial discharge in transformer.Meanwhile,the information fusion method was adopted to recognize results and oil gas features.Experimental results show that,development trend of the same kind of dissolved gas in different discharge models is different,and using information fusion method with the statistical characteristic parameter and the dissolved gases has enough ability to recognize different types of partial discharge in transformers.
出处
《高电压技术》
EI
CAS
CSCD
北大核心
2012年第3期553-559,共7页
High Voltage Engineering
基金
国家自然科学基金青年科学基金(51107155)
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访问学者基金资助项目(2007DA10512711410)~~
关键词
局部放电
模式识别
信息融合
油中溶解气体分析
对应关系
神经网络
partial discharge（PD）
pattern recognition
information fusion
dissolved gas analysis（DGA）
correspondence relationship
neural network
