气体绝缘电器局部放电模式的多特征信息融合决策诊断  

Multiple-characteristics Information-fusion Decision Diagnosis of Partial Discharge Pattern of Gas Insulated Switchgear

作　　者：唐炬[1] 卓然[1] 吴冕之[2] 陶加贵[1] 唐阳[3] 

机构地区：[1]重庆大学输变电装备及系统安全与新技术国家重点实验室,重庆400030 [2]贵阳供电局,贵阳550001 [3]重庆市电力公司北碚供电分公司,重庆400056

出　　处：《高电压技术》2013年第11期2581-2588,共8页High Voltage Engineering

基　　金：国家重点基础研究发展计划(973计划)(2009CB724500);国家自然科学基金(5177181)~~

摘　　要：气体绝缘电器(GIS)局部放电(PD)信息的传统模式识别方法存在识别率低、特征维数过高、信息利用不充分等问题。为此,根据Dempster-Shafer证据理论,将对局部放电的时间解析信号分析模式和局部放电的相位解析信号分析模式的融合决策诊断引入GIS故障分析,实现了对GIS内部绝缘缺陷的有效综合判别。仿真与实验结果表明,该方法能深入挖掘识别中的有效信息,并通过一定的决策规则实现对信息的总体描述,具有良好的拒误诊断能力。此外,基于2种分析模式对4种典型缺陷的综合识别率均高于80%,且互补性较强,在降低特征维数的同时显著提高了识别准确度。Conventional methods of pattern recognition for partial discharge（PD） information in gas insulated switchgear（GIS） have many shortcomings, such as low recognition accuracy, high feature dimension, insufficient utilization of information, and so on. To overcome these shortcomings, on the basis of the Dempster-Shafer evidence theory, we introduced the fusion decision diagnosis of the signal analysis models of time-resolved partial discharge and phase-resolved partial discharge into the fault analysis of GIS, and achieved the comprehensive recognition of insulation defects in GIS. Further simulation and experiments show that,using the proposed method is able to make full use of effective information in recognitions and provide descriptions of all the information by regulations while having a superior ability to avoid misdiagnosis. Analyzing four typical defects shows that the comprehensive recognition rates of the two analysis modes are both above 80 % accompanied with excellent complementary capability. In conclusion, the method reduces the dimension of defect features and improves the recognition accuracy.

关 键 词：局部放电 气体绝缘电器 反向传播神经网络 DS证据理论 模式识别 融合总裁诊断 

分 类 号：TM855[电气工程—高电压与绝缘技术] TP202[自动化与计算机技术—检测技术与自动化装置]