Computer Engineering and Applications ›› 2017, Vol. 53 ›› Issue (4): 25-32.DOI: 10.3778/j.issn.1002-8331.1607-0382

Previous Articles     Next Articles

Optimal particle enhanced exploration particle swarm optimization

TANG Yiling, JIANG Shunliang, YE Famao, XU Qingyong, GE Yun, XU Shaoping   

  1. School of Information Engineering, Nanchang University, Nanchang 330031, China
  • Online:2017-02-15 Published:2017-05-11

最优粒子增强探索粒子群算法

唐祎玲,江顺亮,叶发茂,许庆勇,葛  芸,徐少平   

  1. 南昌大学 信息工程学院,南昌 330031

Abstract: A modified Particle Swarm Optimization(PSO), Optimal particle Enhanced Exploration Particle Swarm Optimization(OEEPSO), is presented with the aim of solving the problem of premature convergence, slow searching speed and low solution accuracy. A new strategy is used to update the velocity and the position of optimal particle. It divides D-dimensions of optimal particle into several groups with each group contains two dimensions. Each group is updated by the position which has the best fitness in four different directions. This strategy expands the searching space around optimal particle and let optimal particle move to the position which is nearer to the optimum solution. This strategy accelerates the searching speed and reaches higher solution accuracy. OEEPSO also proposes a new strategy to avoid local optima. It utilizes the fitness of optimal particle to determine the velocity, so that particle swarm can escape from local optima effectively and find a better optimal particle. OEEPSO has been tested on 6 benchmark functions. Results show that OEEPSO has better performance than many other PSO algorithms in terms of convergence speed, global optimality and solution accuracy.

Key words: particle swarm optimization, optimal particle, enhanced exploration, dimension division, fitness

摘要: 针对粒子群优化算法(Particle Swarm Optimization,PSO)存在收敛速度慢、寻优精度低和早熟收敛的问题,提出一种最优粒子增强探索粒子群算法(Optimal particle Enhanced Exploration Particle Swarm Optimization,OEEPSO)。OEEPSO将最优粒子在空间中的位置信息以二维一组划分,按4种方式计算每二维的适应值,选择适应值最小的方式更新对应维度的速度值和位置值。该策略加强了对最优粒子周围区域的探索,使粒子群能更快地向全局最优解靠近,提高了算法的收敛速度和求解精度。当算法陷入局部最优时,根据群体历史最优解的适应值,动态调整各粒子的速度值和位置值,使算法最终收敛到全局最优解。实验结果表明,OEEPSO具有收敛速度快、求解精度高的特点。

关键词: 粒子群算法, 最优粒子, 增强探索, 维度划分, 适应值