摘要
分析了粒子群优化(PSO)算法的进化式,针对其容易发生早熟、收敛速度慢、后期搜索性能和个体寻优能力降低等缺点,结合遗传算法的思想,提出一种新的混合PSO算法——遗传PSO(GAPSO)。该算法是在PSO算法的更新过程中,对粒子速度引入遗传算法的变异操作,对粒子位置引入遗传算法交叉操作。对速度的变异降低了算法后期因种群过于密集而陷入局部最优的可能,对位置的交叉使得父代中优良个体的基因能够更好地遗传给下一代,从而得到更优、更多样化的后代,加快进化过程,提高了收敛速度和群体搜索性能。选取了其他几种典型的改进PSO算法,从算法执行过程、参数设置及优化性能等方面对各算法进行全面的分析比较,其中对模拟退火PSO算法采用了一种新的可提高算法执行速度的退火方式。最后针对选取的六个Benchmark函数优化问题进行数值仿真实验。仿真结果表明了所提出的遗传PSO算法不但收敛速度加快,而且后期搜索性能提高,能更有效地收敛到全局最优。为了形象地显示粒子的收敛过程,还仿真了GAPSO算法对二维多模态Grie-wank函数的动态寻优过程。
To deal with the slow search speed, premature convergence and lower search performance and individual optimizingability in late stage, this paper proposed a new PSO called genetic PSO. Produced mutation and crossover of GA into velocityand position updating of PSO. The mutation to velocity could reduce the possibility of the algorithm trapping in the local opti-mal because of the over dense of the population in late stage. The crossover to position could make the gene of excellent elderindividuals passed down to the next generation, and by doing so, attained the more excellent and more various next genera-tions, so increased the evolution and search performance of the population. Selected several other typical improved PSO algo-rithms for comparing and analyzing from implementing process, setting of parameters and optimization performance. To simula-ted annealing PSO, proposed a new annealing method which could increase the speed of implementation of the algorithm. Thesimulation experiments were done to the six selected Benchmark functions. The results show that the proposed algorithm notonly speeded up the convergence, but also improved the search performance in late stage and could converge to the global opti-mal solution more efficiently. And lastly, presented the simulation of dynamic optimizing process of genetic PSO to the Grie-wank function so that converging process of the particles could be viewed vividly.
出处
《计算机应用研究》
CSCD
北大核心
2010年第2期453-458,共6页
Application Research of Computers
基金
国家自然科学基金资助项目(60773224)