基于自适应PSO算法的LS-SVM牵引变压器绝缘故障诊断模型  

Insulation fault diagnosis model based on adaptive PSOand LS-SVM for traction transformer

作　　者：方科[1] 黄元亮[1] 刘新东[1] 

机构地区：[1]暨南大学电气自动化研究所,广东珠海519070

出　　处：《电力自动化设备》2011年第3期85-89,共5页Electric Power Automation Equipment

基　　金：国家自然科学基金项目(51007030);铁道部科技研究开发计划课题(2008J002)~~

摘　　要：结合罗杰斯比值法,基于自适应PSO算法和最小二乘支持向量机(LS-SVM),提出一种牵引变压器绝缘故障诊断模型。该模型提出一种自适应PSO算法,即根据群体的收敛程度和个体的适应值来调整惯性权重,加快训练速度,利用该算法迭代求解LS-SVM中出现的矩阵方程,避免矩阵求逆,节省内存;为了快速和准确地区分牵引变压器12种绝缘故障,该模型构建12个自适应PSO的LS-SVM分类器。通过对600组牵引变压器的故障数据的处理表明,所提出的基于自适应PSO优化的LS-SVM算法优于经典SVM算法和标准PSO的LS-SVM算法,收敛速度快,识别精度高。Combined with the Rogers ratio method,an insulation fault diagnosis model based on adaptive PSO and LS-SVM is presented for traction transformer.It adjusts the inertia weights to accelerate the training speed according to the swarm convergence and individual fitness.The matrix of LS-SVM is solved through the adaptive PSO to avoid the inverse solution for memory saving.In order to quickly and accurately distinguish 12 insulation faults,12 LS-SVM classifiers based on adaptive PSO are built.Tests on 600 groups of fault data indicate that,the proposed model has faster convergence velocity and higher accuracy than classical SVM model and LS-SVM model based on standard PSO.

关 键 词：故障诊断 牵引变压器 最小二乘支持向量机 粒子群优化 罗杰斯比值法 多分类 

分 类 号：TB855[一般工业技术—摄影技术] TP277[自动化与计算机技术—检测技术与自动化装置]