Computer Engineering and Applications ›› 2008, Vol. 44 ›› Issue (11): 52-56.

• 理论研究 • Previous Articles     Next Articles

Particle Swarm Optimization algorithm with Varying Population Size

WANG Guang-hui1,2,ZENG Jian-chao1   

  1. 1.Division of System Simulation and Computer Application,Taiyuan University of Science and Technology,Taiyuan 030024,China
    2.Educational Technology Center,Shijiazhuang Command Institute of Chinese People’s Armed Police Force,Shijiazhuang 050061,China
  • Received:2007-07-27 Revised:2007-10-09 Online:2008-04-11 Published:2008-04-11
  • Contact: WANG Guang-hui

一种具有动态群体规模的微粒群算法

王光辉1,2,曾建潮1   

  1. 1.太原科技大学 系统仿真与计算机应用研究所,太原 030024
    2.武警石家庄指挥学院,石家庄 050061
  • 通讯作者: 王光辉

Abstract: Based on the standard Particle Swarm Optimization algorithm(PSO),an improved PSO with Varying Population Size(VPPSO) is proposed.During the evolution of PSO,if the best fitness value(PGBEST) of population is not changed for some generations,the offspring,generated by crossover operator of genetic algorithm,join in the actual population by some deterministic rules.When the population size reaches the maximum size allowed,it is shrinked to the size of starting point by selection operator of genetic algorithm.Through the experiments of four multi-peak test functions with high dimension,the results of simulation show that the rate of convergence and the precision of convergence for VPPSO are better than standard PSO at the high dimension of search space.

Key words: population size, Particle Swarm Optimization(PSO), multi-peak function, high-dimension function

摘要: 以标准微粒群算法PSO为基础,提出了一种改进的群体规模可变的微粒群算法—VPPSO。该方法是在标准PSO的进化过程中,当PGBEST(全局最好值)连续多代不发生变化时,利用遗传算法的杂交机制产生子代,并根据一定的规则加入进化群体中,当群体规模超过允许的最大值时,再通过选择机制,将群体规模收缩到初始时的状态。通过对四个多峰测试函数进行仿真,其结果表明:在高维多峰函数的优化中,VPPSO的收敛率以及收敛精度较标准PSO有很大的提高。

关键词: 群体规模, 微粒群算法, 多峰函数, 高维函数