远离最差解的粒子群优化算法

An Improved Particle Swarm Algorithm Far Away from Worst Solution

针对传统粒子群优化算法容易陷入局部最优、寻优精度低及后期搜索速度慢等缺陷,提出一种参考局部最差解影响的粒子群算法.当算法搜索后期,全局最优解(global best solution,Gbest)无变化时,局部最优解(personal best solution,Pbest)等于Gbest,这时速度靠拢最优方向向量为零,粒子前进的方向只有自身惯性.而本文加入了局部最差(partial worst solution,Pworst)之后的算法使粒子的前进方向不仅受自身惯性的影响,而且可以继续的寻优,从而找到Gbest.算法采用远离全局最差解和局部最差解的思想,对粒子群优化算法的速度更新公式进行改进,并分别测试全局最差解和局部最差解对粒子群优化算法的影响.通过几个典型的测试函数仿真结果表明,改进后的算法在搜索速度、寻优精度、鲁棒性方面较粒子群算法有了显著提高,而且具有跳出局部最优的能力.

A new particle swarm optimization(PSO)algorithm is proposed in reference to worst solution of being local in standard PSO to avoid the problems of being easy to converge to a local minimum,premature convergence and low precision.In the late period of particle swarm optimization and global optimum(Gbest)when there is no change,the local optimum(Pbest)is equal to the Gbest.Then the speed to move closer to the optimal direction vector is zero,and particles go along the direction only by its own inertia.When the proposed algorithm is added to the local minimum(Pworst)algorithm,the particle's forward move is not only affected by its own inertia,but it alsocan continue to find the best,so as to find the global optimal.This algorithm,by applying the idea of the particles avoiding the worst local and global solution,improves the velocity updating formula based on the PSO,and tests respectively the worst global solution and local worst solution on the influence of particle swarm optimization algorithm.Several typical test function simulations show that the proposed algorithm not only has great advantages of convergence property over some other modified PSO algorithms,but is also effective in avoiding being trapped in local optimal solution.