基于PSO-LIBSVM的广域后备保护新算法

广域后备保护利用电网线路IED和母线lED采集各元件相关信息，构造基于PSO—LIBSVM的线路和母线故障识别网络。以IED采集的信息作为输入向量，建立故障识别网络的训练样本矩阵；PSO算法寻找最优LIBSVM运行参数并训练故障识别网络，利用随机故障下的测试样本矩阵对线路和母线识别网络进行测试，以识别电网故障元件IED，进而确定故障元件。通过大量实验仿真了多种信息不准确情况下的故障识别，实验证明基于PSO-LIBSVM的广域后备保护新算法，识别准确率高，具有很好的容错性。...

Full description

Saved in:
Bibliographic Details
Published in电力系统保护与控制 Vol. 41; no. 15; pp. 49 - 58
Main Author 吴浩 李群湛 刘炜
Format Journal Article
LanguageChinese
Published 四川理工学院自动化与电子信息学院，四川 自贡 643000%西南交通大学电气工程学院,四川 成都,610031 2013
西南交通大学电气工程学院，四川 成都 610031
Subjects
PSO—LIBSVM
容错性
广域后备保护
故障识别
母线IED
线路IED
广域后备保护
线路IED
故障识别
fault tolerance performance
bus IED
PSO-LIBSVM
wide area backup protection
母线IED
容错性
line IED
fault identification
Online AccessGet full text
ISSN1674-3415

Cover

More Information
Summary:广域后备保护利用电网线路IED和母线lED采集各元件相关信息，构造基于PSO—LIBSVM的线路和母线故障识别网络。以IED采集的信息作为输入向量，建立故障识别网络的训练样本矩阵；PSO算法寻找最优LIBSVM运行参数并训练故障识别网络，利用随机故障下的测试样本矩阵对线路和母线识别网络进行测试，以识别电网故障元件IED，进而确定故障元件。通过大量实验仿真了多种信息不准确情况下的故障识别，实验证明基于PSO-LIBSVM的广域后备保护新算法，识别准确率高，具有很好的容错性。
Bibliography:WU Hao1,2, LI Qun-zhan1, LIU Wei1 （1. School of Electrical Engineering, Southwest Jiaotong University, Chengdu 610031, China; 2. School of Automation and Electronic Information Engineering, Sichuan University of Science ＆ Engineering, Zigong 643000, China）
41-1401/TM
The related information of different power system element may be collected by line IED and bus IED for wide-area backup protection, then the fault recognition network of line and bus based on PSO-LIBSVM is established and the information collected by IED is taken as input vector for establishing the training sample matrix of the fault recognition network. PSO algorithm is used to search optimal parameter combination to train the fault identification network. The line and bus identification network is tested by the testing sample matrix of random failure, and the fault component IED is identified and the fault components are determined. A large number of experiments are made to simulate the fault identification under a variety of inaccurate informat
ISSN:1674-3415