基于LMD-1(1/2)维谱熵-Elman神经网络输电线路短路故障识别理论与方法  

Power system transmission line short circuit fault recognition theory and method based on LMD-1(1 /2)dimension spectrum entropy- elman neural network

作　　者：蔡明山[1] 蔡双[2] 周军[3] 

机构地区：[1]湖南文理学院电气与信息工程学院,湖南常德415000 [2]常德烟草机械有限责任公司,湖南常德415000 [3]国网江西省电力公司宜春供电分公司,江西宜春336000

出　　处：《湖南科技大学学报（自然科学版）》2015年第1期78-86,共9页Journal of Hunan University of Science And Technology：Natural Science Edition

摘　　要：针对电力系统输电线路故障时短路电流的暂态特征,采用LMD对相模变换后的短路电流进行分解,得到一系列PF分量,然后计算前8个PF分量的1(1/2)维谱熵值作为特征向量,最后将构造的特征向量输入到已训练好的Elman神经网络中进行故障类型识别,并在Matlab平台上建立仿真模型.仿真结果表明,采用的方法能够快速准确地判断出故障类型和故障相;与传统BP网络相比,该方法具有更快的识别速度、更高的识别率,并且识别结果不受过渡电阻、故障位置、相差角等线路参数的影响,因而,实用、有效.Considering the transient characteristics of short-circuit when fault occurred in power system transmission lines, the short-circuit currents are decomposed by applying LMD, and a series of PF components were obtained, then by calculating the 1 （1/2） dimension spectrum entropy of the fore 8 PF components among which as a feature vector which at last be put into the well trained Elman neural network to be realized fault recognition. A simulation model was established by Matlab, and the simulating results show that the method fast and accurately judge the fault type and fault phase, compared with the traditional BP network, the method has faster recognition speed, higher recognition rate, and also the recognition result is not influenced by the line parameters like transition resistance, fault location and phase angle etc. , so it is effective and practical.

关 键 词：输电线路 故障识别 LMD分解 1(1/2)维谱熵 ELMAN神经网络 

分 类 号：TM713[电气工程—电力系统及自动化]