基于混沌优化粒子群BP神经网络的电力变压器故障诊断  

Fault diagnosis of power transformers based on chaos particle swarm optimization BP neural network

作　　者：公茂法[1] 柳岩妮 王来河 宋保业[1] 钟文强 

机构地区：[1]山东科技大学,山东青岛266590 [2]国网山东平度市供电公司,山东平度266700

出　　处：《电测与仪表》2016年第15期13-16,32,共5页Electrical Measurement & Instrumentation

基　　金：山东省自然科学基金项目(ZR2012EEM021)

摘　　要：针对电力变压器故障诊断问题,提出了一种基于混沌(Chaos)优化的粒子群(Particle Swarm Optimization)BP神经网络算法。该算法将混沌、粒子群和BP神经网络相结合,通过混沌粒子群算法寻优,得到BP神经网络的最优权值和阈值初始值,然后进行网络训练和测试。利用了混沌算法的遍历性和对初始值敏感的特点,对粒子群算法进行了参数优化,引入了早熟判断机制,并在早熟状态时进行了混沌扰动,使算法后期不易陷入局部最优。通过实例训练与测试表明,CPSO-BP神经网络算法在变压器故障诊断方面有较好的效果。In order to solve the fault diagnostic problems of power transformers,this paper proposes a BP neural network algorithm based on chaos particle swarm optimization. This algorithm combines chaos,particle swarm and the BP neural network,and obtains the optimal weights and the initial threshold value of BP neural network by using chaos particle swarm optimization algorithm,and then,it takes the network training and testing. The algorithm takes advantage of ergodicity and sensitivity of initial value of chaos to optimize the parameters of particle swarm algorithm,and introduces the precocious judgment mechanism of local convergence,besides that,in order to avoid the algorithm easily falling into local convergence,the algorithm takes chaotic disturbance at the precocious state. The training and testing examples suggest that CPSO-BP neural network algorithm has better effect in transformers fault diagnosis.

关 键 词：混沌 粒子群 BP神经网络 变压器 故障诊断 

分 类 号：TM93[电气工程—电力电子与电力传动]