摘要
在两种微粒群算法分析的基础上,针对算法存在局部最优和后期振荡的现象,提出一种改进自适应微粒群算法.新算法引入概率突跳因子改变了原算法中微粒的速度更新公式,引入模拟退火接受准则抑制了概率突跳的不可控制性.典型函数寻优结果表明,新算法能很快地收敛到全局最优解,大幅度降低了达到最优值所需要的迭代数,同时提高了算法的收敛率和收敛精度,在跳出局部搜索的能力上远优于标准微粒群算法和自适应微粒群算法,稳定性好.
On the basis of analyzing two particle swarm optimization（PSO） algorithms, the standard PSO（SPSO） and self-adapting PSO（SAPSO）, a modified adapting PSO（MAPSO） algorithm is proposed to solve the problem that PSO may trap to local optimum and fluctuation during later period. In this algorithm, the probabilistic leap factor is introduced to modify the velocity updating and the acceptable rule of simulated annealing is applied to restrain the uncontrollability of probabilistic leap. The results of typical optimization show that this algorithm has better accuracy and convergence rate as well as fewer iteration numbers in approaching the global optimization than SPSO and SAPSO algorithms. This algorithm is also superior to SPSO and SAPSO algorithms in stability and ability of breaking off local search.
出处
《控制与决策》
EI
CSCD
北大核心
2009年第4期617-620,627,共5页
Control and Decision
基金
国家自然科学基金项目(60774087)
航天支撑技术基金项目(0711205)
关键词
微粒群优化
模拟退火
自适应
概率突跳
PSO
Simulated annealing
Self-adapting
Probabilistic leap
