基于粒子群-遗传的混合优化算法

Hybrid optimization algorithms based on particle swarm optimization and genetic algorithm

提出了一种基于实数编码的粒子群优化和遗传算法的混合优化算法,该算法首先由粒子群优化进化一定代数后,将最优的M个粒子保留,去掉适应度较差的pop_size-M个粒子。然后以这最优的M个粒子的位置值为基础,选择复制得到pop_size-M个个体,并进行交叉、变异等遗传算法运算。最后将保留的M个粒子位置值与遗传算法进化得到新的pop_size-M个体合并形成新的粒子种群,进行下一代进化运算。该算法在进化过程中能进行多次信息交换,使两种算法互补性得到更充分的发挥。通过5个函数优化实例与其他多种算法的对比研究,表明该算法收敛性能好,运算速度快,优化能力强。此外,还研究了最优粒子保留规模M以及粒子群优化进化较少代数规模对算法性能的影响。

This paper develops a hybrid optimization algorithms based on particle swarm optimization（PSO） and genetic algorithm（GA）.Firstly,the population are evolved a certainty generations by PSO and the best M particles are retained while the other pop-size-M particles are removed.Secondly,generate pop-size-M new individuals by implementing selection,crossover and mutation operators of GA according to the remaining best M particles.Finally,put the pop-size-M new individuals into the remaining best M particles to form new population for next generation.The algorithm can exchange information several times during the evolvement process,so that the complement of two algorithms can be more fully exploited.The proposed method is used to deal with 5 functions optimization problems,and the results obtained are compared with existent bibliography,showing an improvement over the published methods.Furthermore,this paper studies the impact of M scale on the algorithm performance.