基于改进粒子群算法的变压器故障诊断研究  

Research of Transformer Fault Diagnosis Based on Improved Particle Swarm Algorithm

作　　者：李凌[1] 倪远平[1] 孙婧雅[1] 

机构地区：[1]昆明理工大学信息工程与自动化学院,昆明650051

出　　处：《化工自动化及仪表》2010年第7期43-46,共4页Control and Instruments in Chemical Industry

基　　金：国家自然科学基金资助项目(30860055)

摘　　要：针对基本粒子群算法存在收敛慢、易陷入局部极值的缺点,分析了粒子群算法中惯性权重和加速因子的作用,对其作了修改,并用改进后的粒子群算法训练神经网络,应用在变压器故障诊断上。仿真结果表明:改进后的粒子群算法迭代次数少,收敛速度比改进的BP算法快,可以对变压器的故障类型进行区分。Because the standard particle swarm optimization algorithm had some shortcomings of slow convergence and easy to fall into the local extreme value,the effects of inertia weight and accelerating factor in particle swarm optimization were analyzed and the inertia weight and accelerating factor were modified.The improved particle swarm algorithm was used to train neural network and applied to transformer fault diagnosis.Simulation results show that the iterations are fewer,convergence rate is faster than the improved BP algorithm,and the transformer fault types can be distinguished.

关 键 词：改进粒子群算法 变压器故障诊断 神经网络 收敛速度 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]