一种基于运动方向变异的混合粒子群算法研究  

A Hybrid Particle Swarm Optimization Based on the Direction Variation

作　　者：陈明杰[1] 郭少锋[1] 张旻[1] 

机构地区：[1]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001

出　　处：《山东科技大学学报（自然科学版）》2013年第6期84-88,共5页Journal of Shandong University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(51079033);中央高校基本科研业务费专项资金项目(HEUCF100430)

摘　　要：针对粒子群算法搜索精度不高、易早熟收敛、搜索后期多样性下降快等问题,提出一种基于运动方向变异的混合改进粒子群算法。该算法通过改变部分粒子的运动方向增加种群多样性,扩大粒子的搜索范围;利用非线性减小惯性权重的方法增加搜索后期的精度;用线性地增大和减小两个学习因子来平衡搜索的范围和精度,使得在搜索前期能够迅速定位到全局最优点附近,在搜索后期能够收敛到全局最优点。将该方法应用于函数优化中,仿真结果表明,该算法能够使粒子均匀分布在最优值空间范围内,调整和平衡粒子的全局搜索和局部精细搜索能力,同时能延缓粒子多样性的下降速度,使粒子能够跳出局部最优值。A hybrid improved particle swarm algorithm was proposed to solve the problem of low precision, prema- ture convergence and fast fall of diversity of particle swarm algorithm. It increased population diversity to expand the search scope by chauging the movement direction of some particles, and improved the accuracy by reducing non- linear inertia weight. In additiou, it balanced the search range and accuracy by linear increase and decrease of the two sludy factors, which made it position nearby of global optimal values in early stage, and converged to the global optimal later. The simulation results show that the hybrid improved particle swarm algorithm can achieve the fol- lowing：Overcome the defect of low space ergodicity, delay the speed of diversity decreasing, jump out of the local optimum value with higher degree of accuracy, faster speed of optimization, and higher convergence ratio.

关 键 词：运动方向变异 混合改进粒子群 惯性权值 收敛性 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]