带审敛因子的变邻域粒子群算法

Particle swarm optimization and variable neighborhood search algorithm with convergence criterions

针对基本粒子群算法在求解高维空间中的复杂多峰函数时容易发生早熟收敛而陷入局部最优的问题,汲取变邻域搜索算法全局搜索的优势,提出了带审敛因子的变邻域粒子群算法.首先由基本粒子群的快速搜索能力得到较优的群体;然后通过审敛因子判断发生早熟收敛的粒子,并利用变邻域搜索算法的全局搜索能力对陷入早熟收敛的粒子进行优化,从而得到全局最优.相关实验表明,带审敛因子的粒子群算法的性能较常规粒子群算法更加优越.

For the complex multi-peaks function with high dimension, the particle swarm optimization and variable neighborhood search algorithm with convergence criterions（VNS-PSO-CC） is proposed on the basis of analyzing the problem of premature. This method combines the particle swarm optimization（PSO） with the global search ability of variable neighborhood search（VNS） algorithm, and adds the convergence criterions. Firstly, the preferable swarm is obtained by using the fast searching ability of PSO algorithm. Furthermore, the premature swarm, which is estimated by convergence criterions, is optimized by using VNS algorithm. Finally, experimental results show that the performance of VNS-PSO-CC algorithm is superior to the traditional PSO algorithm.