改进微粒群算法在光子晶体优化中的应用  

Improved Particle Swarm Optimization and Its Application in Photonic Crystal

作　　者：曾渊[1] 李源[1] 许家栋[1] 

机构地区：[1]西北工业大学电子信息学院,陕西西安710072

出　　处：《计算机仿真》2008年第3期202-205,共4页Computer Simulation

基　　金：陕西省自然科学基础研究计划(2006F15);西北工业大学科技创新基金(2006CR11)

摘　　要：针对标准微粒群算法无法合理控制全局搜索和局部开发之间的关系,容易出现早熟收敛和全局收敛放慢的现象,提出了一种基于吸引力排斥力平衡机制的改进微粒群算法。改进算法将优化过程分为三个阶段,设定了每个阶段的目标,以此为指导来分别调整引力斥力大小和极优值传播速度,有重点地进行全局搜索或局部开发,以达到提高优化效率的目的。采用四个典型测试函数对改进算法进行了测试,并将该算法应用在光子晶体带隙优化设计中,实验结果表明,改进微粒群算法很好地避免了早熟收敛和全局收敛放慢的现象,相比标准算法具有较高的可靠性和稳定性,是一种高效的优化算法。In order to balance the relationship of globe exploration and local exploitation efficiently, and avoid the premature convergence and slowdown convergence phenomenon, the improved Particle Swarm Optimization （ IPSO）, based on balance mechanism of attraction and repulsion, is proposed. Optimization process is divided into three stages and each stage has its own goals. The paper uses it as guidance to control the value of attraction and repulsion and propagation velocity of excellent value, and focuses on global search or local development in order to enhance the effi- ciency of algorithm. The experiment results of four benchmark function and application in photonic crystal optimiza- tion indicate that IPSO can well avoid the phenomenon of premature convergence and slowdown convergence, compared to PSO, so the IPSO is a highly reliable, stable and efficient algorithm.

关 键 词：平衡机制 传播参数 多阶段 微粒群优化算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]