基于GMAPM和SOM-LVQ-ANN的输电线路故障综合识别方法  

Comprehensive recognition method of transmission line faults based on GMAPM and SOM-LVQ-ANN

作　　者：孙晓明[1,2] 秦亮 刘涤尘[2] SUN Xiaoming;QIN Liang;LIU Dichen(Department of Electrical Engineering,Chongqing Water Resources and Electric Engineering College,Chongqing 402160,China;School of Electrical Engineering and Automation,Wuhan University,Wuhan 430072,China)

机构地区：[1]重庆水利电力职业技术学院电气工程系,重庆永川402160 [2]武汉大学电气与自动化学院,湖北武汉430072

出　　处：《武汉大学学报（工学版）》2019年第12期1079-1090,1105,共13页Engineering Journal of Wuhan University

基　　金：国家自然科学基金青年科学基金项目(编号:51207115);重庆市教育委员会科学技术研究项目(编号:KJ1603605);永川区自然科学基金计划项目(编号:Ycstc,2016nc3001)

摘　　要：现有输电线路故障识别方法大多不能同时识别输电线路的低/高阻抗故障和发展性故障以及电力系统的异常工况(包括低频振荡、铁磁谐振和PT/CT饱和等)和此工况下的故障,故不能满足除继电保护领域外的继电保护测试领域及大电网事故分析和预警防御领域的新的应用需求.因此,提出一种基于广义改进自适应Prony方法(generalized modified adaptive Prony method,GMAPM)和自组织映射-学习向量量化-人工神经网络(self-organizing mapping-learning vector quantization-artificial neural network,SOM-LVQ-ANN)的输电线路故障综合识别方法,以期能同时识别以上输电线路故障和电力系统异常工况及异常工况下的故障.其中,作为信息提取环节的GMAPM实现了多路信号的并行处理和同时分析,作为特征识别环节的SOM-LVQ-ANN继承了SOM-ANN的强自主学习能力和泛化能力以及LVQ-ANN可预先指定故障类型且便于类型编码和拓展的优点.仿真实验结果初步验证了本方法的优良性能.Most of the current recognition methods of transmission line faults are unable to simultaneously recognize the low/high impedance faults and the evolved faults of transmission lines together with the abnormal working conditions of power system(including low frequency oscillation,ferromagnetic resonance and PT/CT saturation,etc.)and the faults under these conditions;therefore,except the protective relay field,the current recognition methods can not meet the new application requirements from the protective relay testing field and the fault analysis and warning/defense field of bulk power grid.Due to this,this paper proposes a comprehensive recognition method of trans-mission line faults based on the generalized modified adaptive Prony method(GMAPM)and the self-organizing mapping-learning vector quantization-artificial neural network(SOM-LVQ-ANN),in order to be able to recognize the transmission line faults,the abnormal working conditions of power system and the faults under these conditions at the same time.Thereinto,GMAPM,the information extraction part,realizes the parallel processing and simultaneous analysis of multi-channel signals;SOM-LVQ-ANN,the feature recognition part,inherits the strong independent learning ability and generalization ability of SOM-ANN and the advantages of preassigning,coding and expanding fault types of LVQ-ANN.The simulation experimental result has preliminarily verified the excellent performance of the method.

关 键 词：输电线路故障综合识别方法 广义改进自适应Prony方法 自组织映射-学习向量量化-人工神经网络 

分 类 号：TM726[电气工程—电力系统及自动化]