Abstract: Information fusion algorithm based on Particle Swarm Optimization-based Support Vector Machine（PSO-SVM） is proposed as fault element discriminant for wide area backup protection. Related information of multiple nodes need to be collected by wide-area backup protection to determine a regional power grid fault element. In this paper, three kinds of information of the line fault directional component, the measurement element of distance protection paragraph II and the status of main protection are selected. The state information matrix for all fault is made as the training sample to train PSO-SVM network, then the state information vector of some random failure is used as the testing sample, and a large number of experiments are made to simulate different fault identification results with inaccurate information. The results of experiments show that wide area backup protection based on PSO-SVM has good ability of fault-tolerance and high accuracy.

Key words: particle swarm optimization, support vector machine, information fusion, wide-area back-up protection, fault identification

摘要： 提出了基于粒子群优化支持向量机（PSO-SVM）的信息融合算法进行广域后备保护故障元件判别。广域后备保护需要采集多个节点相关信息以判别电网某区域的故障元件，选取了线路故障方向元件、线路距离II段元件和主保护动作状态三类信息。利用确定故障下的状态信息矩阵作为PSO-SVM的训练样本，再用随机故障时的元件状态信息矩阵作为测试样本。通过大量实验模拟了多种信息不确定情况下的故障判别结果。实验结果表明，基于PSO-SVM的保护算法具有很好的容错和正判能力。

关键词: 粒子群优化, 支持向量机, 信息融合, 广域后备保护, 故障判别