基于空间缩放和吸引子的粒子群优化算法

摘要

To control particles to fly inside search space and deal with the problems of slow convergence speed and premature convergence of particle swarm optimization (PSO) algorithm, this paper studies the movement of particles and stability analysis of canonical PSO algorithm and proposes an improved PSO algorithm, called PSO with search space zoomed factor and attractor (SzAPSO), where search space zoomed factor is a key parameter to control the original search space to zoom in and out, which is benefit to retain the connection of particles' position, reduce the subjective interference, and enforce the ability of global search, and attractor is a weighted average of global best and personal best for the normal particles except the global-best particle,which utilizes known information to enhance the power of local search and escaping from an inferior local optimum. SzAPSO is not only a kind of boundary condition, but also an effective PSO algorithm. Experimental studies show that SzAPSO algorithm proposed in this paper is more effective to do with errant particles, furthermore, improves greatly the convergence speed and accuracy, and obtains the admirable optimization results with smaller population size and evolution generations independent of the problem dimension and the location of the global optimum with respect to search space boundary.%为解决粒子群优化(Particle Swarm Optimization,PSO)算法中粒子越界、算法进化后期收敛速度慢和早熟收敛的问题,通过分析PSO算法中粒子运动行为和算法稳定性,提出了一种基于空间缩放和吸引子的粒子群优化(PSO with search space zoomed factor and attractor,SzAPSO)算法.该算法利用对搜索空间进行缩放的边界变异策略有效控制了粒子搜索范围,保证了算法全局探测能力;算法中吸引子的引入增加了感兴趣区域的粒子密度,提高了算法局部开发能力.实验结果表明,SzAPSO算法收敛速度快、精度高,且具有较好的鲁棒性.