摘要
针对标准粒子群算法在处理复杂优化问题时易出现收敛速度慢和陷入局部最优的问题,提出了一种自适应进化模型的粒子群优化算法。通过设定的阈值limit将种群进化状态划分为正常状态和"早熟"状态,当种群全局最优位置信息连续超过limit次没有更新时,认为算法处于"早熟"状态,此时对种群的个体最优位置进行反向学习,帮助算法逃离局部最优,并采用新的进化模型;否则视为正常进化状态,并采用标准粒子群进化模型。8个基准测试函数的仿真结果表明,该算法与一些其它改进粒子群算法如FIPS、CLPSO、MPSO-SFLA算法相比,在全局寻优能力、收敛速度和收敛精度方面都具有明显的优势。
As standard particle swarm optimization algorithm had some shortcomings, such as converging slowly and getting trapped in the local minima, a new improved PSO algorithm based on adaptive evolution was proposed. The algorithm's population evolution state was divided into “normal and premature” by setting the threshold value limit. When population's global optimal position was not updated continuously for more than limit times, the algorithm was considered to be in the “premature” state. At this point, the individuals' optimal position adopted opposition-based learning strategy which helped the algorithm to escape from local optimum, and the algorithm adopted the improved evolution model. Otherwise, it was considered to be in “normal” evolution state, and the standard model was adopted. The simulation experiment results on eight classical benchmark functions showed that the capacity of searching optimal solution, convergence speed and convergence accuracy of the AEMPSO was better than some of the common improved particle swarm optimizations, such as FIPS, CLPSO and MPSO-SFLA.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第8期2901-2906,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(61301232)
关键词
粒子群优化算法
自适应进化
反向学习
快速收敛
局部最优
particle swarm optimization
adaptive evolution
opposition-based learning
fast convergence
local optima