首页 | 本学科首页   官方微博 | 高级检索  
     

基于粒子群优化神经网络的变压器故障诊断
引用本文:王晓霞,王涛.基于粒子群优化神经网络的变压器故障诊断[J].高电压技术,2008,34(11):2362-2367.
作者姓名:王晓霞  王涛
作者单位:1. 华北电力大学计算机科学与技术学院,保定,071003
2. 华北电力大学数理学院,保定,071003
摘    要:为克服电气分析应用中误差反向传播(BP)神经网络存在的不足,提出了一种利用改进粒子群算法优化神经网络的变压器故障诊断新方法。该法的惯性权重自适应调整,以平衡局部和全局搜索能力;收缩因子加快算法的收敛速度,有利于更快地收敛于全局最优解。利用改进的粒子群算法优化神经网络参数,并结合BP算法训练网络可有效地克服常规BP算法训练网络权值和阈值收敛速度慢、易陷入局部极小和遗传算法独立训练神经网络速度缓慢等缺点。最后,进行变压器故障实例分析的仿真结果表明,该算法具有较快的收敛速度和较高的诊断准确度,证实了该方法的正确性和有效性。

关 键 词:粒子群优化算法  BP算法  神经网络  变压器  故障诊断  仿真

Power Transformer Fault Diagnosis Based on Neural Network Evolved by Particle Swarm Optimization
WANG Xiao-xia,WANG Tao.Power Transformer Fault Diagnosis Based on Neural Network Evolved by Particle Swarm Optimization[J].High Voltage Engineering,2008,34(11):2362-2367.
Authors:WANG Xiao-xia  WANG Tao
Affiliation:1. School of Computer Science & Technology,North China Electric Power University,Baoding 071003,China; 2. School of Mathematics & Physics,North China Electric Power University,Baoding 071003,China)
Abstract:Power transformer is one of the most important equipments in power network. Its normal operation is the basis of the power supply and normal social life. So it is valuable to discover the incipient fault accurately and timely. This paper proposes a new power transformer fault diagnostic method using neural network evolved by modified particle swarm optimization (modified PSO) algorithm in order to overcome the problem of premature convergence observed in many applications of error back propagation (BP) algorithm and enhance the fault diagnostic ability of conventional dissolved gas-in-oil analysis in power transformer. In the modified PSO algorithm,the inertia weight is adjusted adaptively in order to balance and reconcile the global and local searching capability. The convergence can be accelerated by setting the compression factor of the modified PSO algorithm reasonably,which may benefit to find the global optimal solution quickly. Firstly,the modified PSO algorithm is used to optimize the original parameter of the neural network,then the gradient descent algorithm is used to train the neural network. Defects of conventional BP algorithm,i.e. the slow convergence of weight and threshold learning,premature result,and the slow training speed of GA,are settled by the algorithm. Finally,Simulation results of power transformer fault diagnosis show that both convergence speed and diagnosis accuracy are improved to some extent. That shows the correctness and validity of this method in power transformer fault diagnosis by dissolved gas-in-oil analysis.
Keywords:particle swarm optimization algorithm  BP algorithm  neural network  power transformer  fault diagnosis  simulation
本文献已被 CNKI 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号