摘要
基于正交试验设计的最优性以及微粒群中微粒的记忆特征,提出了一种新型的微粒群算法——正交微粒群算法。其主要思想是:利用正交设计的方法产生初始微粒群,以便粒子能够均匀分布在整个解空间上;充分利用微粒的记忆能力,对微粒群进行更新,从而达到对可行解空间进行开发和探索的目的。将该算法应用于四个常见的测试函数,试验结果表明本算法的性能比较优越,并且具有很强的并行性和较大的灵活性。最后,讨论了不同的初始速度和扰动对算法性能的影响。
A new algorithm based on the optimality of orthogonal experimental design method and the abilities of memory in particles was proposed, which is called Orthogonal Particle Swarm Optimization (OPSO). Its characteristic is that initial particles of particle swarm are generated by orthogonal experimental design, so that these particles can be scattered uniformly over the feasible solution space and the particle swarm of the next generation is generated by means of memory. So the search space could be explored and exploited efficaciously. The OPSO was tested on four benchmark functions. The experimental results illustrate that the OPSO has the potential to achieve faster convergence and to find a better solution and has strong parallel characters and flexible features. In the end, the performance of the new algorithm was discussed caused by different settings of initial velocity and disturbance.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2005年第12期2908-2911,共4页
Journal of System Simulation
基金
国家自然科学基金(60073053和60133010)
河南省自然科学基金(0511013700)
河南省教育厅自然科学基金(2000110019)
河南省高校青年骨干教师计划基金资助。
关键词
微粒群
微粒群算法
函数优化
试验设计
正交设计
particle swarm
particle swarm algorithm
function optimization
experimental design
orthogonal design