多阶段多模型的改进微粒群优化算法  

Modified particle swarm optimization based on multi-stages & multi-models

作　　者：赵嘉[1] 孙辉[1] 

机构地区：[1]南昌工程学院信息工程学院,南昌330099

出　　处：《计算机工程与应用》2010年第33期32-35,共4页Computer Engineering and Applications

基　　金：国家自然科学基金No.50539020;江西省自然科学基金No.2009GZW0020;江西省教育厅科技项目No.GJJ09365;No.GJJ10630;南昌工程学院青年基金-科技项目No.2010KJ015;No.2010KJ018~~

摘　　要：针对微粒群优化算法在解决复杂优化问题时易于出现早熟收敛现象,提出了一种多阶段多模型的改进微粒群优化算法。考虑寻优不同阶段的开发与探测能力需求的差异,算法将寻优过程分成3个阶段,各阶段采用不同的模型进行进化。第一阶段利用标准微粒群优化算法发现局部极值的邻域;第二阶段利用CognitionOnly模型快速找到局部极值点,提高寻优效率;第三阶段,提出了一种改进的进化模型,利于粒子快速跳出局部极值点,寻找到全局最优点。4种复杂测试函数的实验结果表明:该算法比标准微粒群优化算法(PSO)和基于不同进化模型的两群优化算法(TSE-PSO)更容易找到全局最优解,相比两群微粒群优化算法,还能在一定程度上提高优化效率。Modified particle swarm optimization based on multi-stages ＆ multi-models is advanced to solve complex problems since premature convergence phenomena are common in current particle swarm optimization.This algorithm process includes three stages and various models considering development and exploration diversity in different steps.Firstly, the local extremum neighborhood is obtained through PSO.Secondly, local extremum is quickly searched out with Cognition Only to improve evolution efficiency.Lastly,the most optimal solution is sought through the improved evolution model to avoid local extremum.Results of four complex function tests show that the proposed algorithm is easier to get the optimal solution than PSO and TSE-PSO in efficiency and performance.

关 键 词：微粒群优化算法 多模型 多阶段 优化 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]