一种自适应粒子群算法的小波神经网络优化  

A Wavelet Neural Network Optimization Method Based on Variable-Weight Particle Swarm Optimization

作　　者：刘志勇[1] 王小红[2] LIU Zhiyong;WANG Xiaohong(Xianyang Vocational and Technical College,Xianyang 712000,China;Shaanxi Provincial Party School of the CPC,Xi’an 710061,China)

机构地区：[1]咸阳职业技术学院,陕西咸阳712000 [2]陕西省委党校,陕西西安710061

出　　处：《机械与电子》2021年第8期8-12,共5页Machinery & Electronics

基　　金：咸阳市科技计划项目(2018K02-11)。

摘　　要：针对传统粒子群算法易陷入局部最优解的问题,提出了一种变权重粒子群算法。该算法通过引入交叉权重因子和粒子个体状态最优权值,对传统粒子群算法进行了优化,使粒子在移动过程中利用更多的信息来调整各自的移动方向,扩大粒子在运动过程中的自我认知范围,提高了粒子群算法的收敛精度和收敛速度。最后,利用改进的变权重粒子群算法对小波神经网络控制器进行优化,有效地验证了变权重粒子群算法的精确性。A variable-weight particle swarm optimization(VWPSO)algorithm is proposed to overcome the shortcoming of the traditional particle swarm optimization(PSO)algorithm,which is easy to fall into the local optimal solution.In the proposed strategy,the weight factor and the optimal weight of individual state are introduced to adjust the moving direction of the particles via the more information,and the self awareness range of particles in the process of motion is expanded.Moreover,the convergence accuracy and convergence speed of particle swarm optimization algorithm are improved.Finally,the VWPSO algorithm is used to optimize the wavelet neural network controller,which verifies the accuracy of the VWPSO algorithm effectively.

关 键 词：变权重 粒子群算法 全局最优 小波神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]