基于串联协作的多方法协作优化方法  

Multimethod Collaborative Optimization Algorithm Based on Serial Collaboration

作　　者：郭忠全[1] 罗文彩[1] 陈小前[1] 

机构地区：[1]国防科技大学航天与材料工程学院,湖南长沙410073

出　　处：《计算机仿真》2006年第8期170-173,共4页Computer Simulation

基　　金：国家自然科学基金(10302031);国家高技术发展规划(863计划)(2004AA721051)

摘　　要：为更好地求解复杂优化问题的全局最优解,提出了基于串联协作的多方法协作优化方法。它采用串联方式组织各个优化方法之间的协作。各个优化方法在每次协作中进行若干次优化,每一优化方法在优化过程中及优化结果中包含的优化信息被用于其它优化方法中,以提高整体优化性能。选择了遗传算法、模式搜索法和Powe ll法等三个直接优化方法进行串联协作组成的多方法协作优化方法的设计,并对其优化特性进行了深入的理论分析。最后,用三个复杂多维函数对算法性能进行了测试。计算实例表明,基于串联协作的多方法协作优化方法取得了优于单个优化方法的全局最优特性。Muhimethod collaborative optimization algorithm based on serial collaboration is advanced to get better global optima of complex optimization problem. It uses serial structure to collaborate different kinds of optimization methods. Each optimization method is executed several times in one collaboration. Optimal message of each optimization method is used to others to improve the whole optimization performance. A multimethod collaborative optimization algorithm based on serial collaboration is designed here, which is composed of genetic algorithm, pattern search method and Powell＇s method. The optimal characteristic of this algorithm is discussed as well. This algorithm is tested by three muhidimensional functions finally. Results show that the multimethod collaborative optimization algorithm based on serial collaboration obtains better global optimization performance than single optimization method.

关 键 词：多方法协作优化方法 串联协作 遗传算法 模式搜索法 鲍威尔法 

分 类 号：TP202.7[自动化与计算机技术—检测技术与自动化装置]