摘要
提出了一种改进的多群协作粒子群优化算法,该算法整个种群采用主从模式,分为一个主群和多个从群,多个从群粒子统一地进行初始化操作,从而避免了多个粒子群重复搜索现象。同时,算法采取了一种扰动策略,即当前全局最优解在扰动因子的迭代周期内保持不变时,就重置粒子的速度,迫使粒子群摆脱局部极小。该算法不仅增加了种群的多样性,扩大了搜索范围,而且还改善整个种群易陷入局部极小值的缺陷。通过9个基准函数进行测试,实验结果表明,IMCPSO与MCPSO算法相比具有明显的优越性。
A modified multi-swarm cooperative particle swarm optimizer is proposed. The whole population is divided into a master swarm and several slave swarms. So as to avoid the repeated search multiple particle swarm phenomenon , this algorithm initializes particles of slave swarms uniformly. Meanwhile this algorithm adopts a perturbation strategy. Namely , just reset the particle′s velocity for escaping from local minimum, when the current global optimal solution remaines unchanged in the disturbance factor′s iterative cycle. This algorithm not only increases the diversity of population to expand the search scope , but also improves the defects of the whole population easily being fall into local minimum value. The experimental results indicate that IMCPSO has better superiority than MCPSO and PSO by testing 9 benchmark functions.
出处
《微型机与应用》
2014年第15期72-75,共4页
Microcomputer & Its Applications
基金
芬兰科学院基金项目(135225)
上海海事大学研究生创新基金项目(GK2013084)
关键词
多群协作
粒子群优化
函数优化
multi-swarm cooperative
particle swarm optimization
function optimization
