摘要
针对传统粒子群优化(PSO)算法通过单群优化,存在着精度较低、易陷入局部最优解等缺点,提出一种可适性群集变动的微粒算法(ADCPSO)。此算法将依据收敛公式的数值大小,判断粒子群收敛程度,从而动态地调适粒子群群集大小,以提高种群的多样性,有效地避免提早收敛等问题。通过与其他8种粒子群优化算法在CEC'2010标准函数下的仿真测试结果表明:ADCPSO算法凭借着简明算法结构,在寻优能力和算法精度上表现出明显的优势,体现出了较好的应用前景。
The original Particle Swarm Optimization （PSO） through single swarm optimization will cause low precision, easily falls into local minima and other shortcomings. This paper presented a variation on the original PSO algorithm, called Adaptive Dynamic Cluster Particle Swarm Optimization （ ADCPSO）. Through multiple-swarm cluster-searching and a criterion that contains a formula, the convergence of particle swarms was measureed, and the cluster size of particle swarms was dynamically adjusted. The method not only can improve the population diversity of particle swarms but also can effectively adjust the matter with premature convergence. ADCPSO was comprehensively evaluated on benchmark functions in CEC＇2010, and compared with other eight kinds of particle swarm optimization algorithms. The results show that ADCPSO substantially enhances the performance of convergence speed, global optimality search and solution precision. The paper shows that the ADCPSO with concise algorithm structure has superior ability of optimization and better application prospect.
出处
《计算机应用》
CSCD
北大核心
2013年第A02期91-97,104,共8页
journal of Computer Applications
基金
重庆科技学院第八届大学生科技创新训练计划项目(2013064)
关键词
粒子群优化
动态分群优化
可适性
收敛公式
CEC'2010标准测试方程
Particle Swarm Optimization （PSO）
dynamic multi-swarm particle optimizer
adaptive
convergence formula
CEC＇2010 benchmark function
作者简介
杨超（1991-），男，湖南株洲人，主要研究方向：计算智能；通信作者电子邮箱youngchaol@sina．com
李宗勳（1987-），男，台湾桃园人，硕士研究生，主要研究方向：计算智能、图像编码；
莊尧棠（1953-），男，台湾彰化人，教授，博士生导师，博士，主要研究方向：计算智能、语音识别。
