【作者】 王丹；

【导师】 孙越泓；

【作者基本信息】 南京师范大学 ， 计算数学， 2019， 硕士

【摘要】 许多实际的优化问题都包含不等式和等式约束.在过去的几十年中,许多学者开发了一些约束处理技术,并将其和进化算法相结合用于求解带约束的优化问题.根据没有免费午餐定理,单个约束处理技术不可能在每个问题上都优于其他所有技术.受可行搜索空间与整个搜索空间的比值、问题的多模态、选择的进化算法、搜索过程的全局探索或局部开发等因素的影响,不同的约束处理方法在搜索过程的不同阶段起着不同的作用.因此,Mallipeddi等人提出一种约束处理集成技术求解带约束的优化问题.本文对Mallipeddi提出的约束处理集成技术进行改进,提出两种新的约束处理集成技术,并将这两种集成技术分别与人工蜂群算法和差分进化算法结合,用于求解带约束的优化问题.第一个算法是基于约束集成技术的人工蜂群算法.在Mallipeddi提出的约束集成技术的基础上,增加了一种约束处理技术.将五种约束处理技术组合成两种约束集成,分别用在人工蜂群算法的雇佣蜂阶段和观察蜂阶段.在CEC 2017的28个基准测试函数和四个经典的工程设计问题上对新算法的性能进行测试.实验结果表明,该算法可以有效地提高优化问题的解的精度.与其他几种经典的优化算法相比,新算法具有非常强的竞争力.第二个算法是基于新约束集成的差分进化算法.在产生新个体的阶段,算法采用三种不同的突变策略.利用不同的约束处理技术对新个体进行选择,并引入局部搜索,增强算法局部寻优能力.该算法在CEC 2017的28个基准函数上进行数值实验,并且与其他几种经典的算法进行比较,实验结果显示,新算法在求解精度上表现较好.更多还原

【Abstract】 Many practical optimization problems involve inequality and equal-ity constraints.Over the past several decades,several constraint pro-cessing techniques have been developed for evolutionary algorithm-s(EAs)by many scholars.According to the no free lunch theorem,a single constraint processing technology can’t be superior to all oth-er techniques in every problem.Influenced by many factors,such as the ratio between feasible region and the entire search space,the multi In other words-modality of the problem,the selection of EA,and the overall exploration/local exploitation phase,the various con-straint handling techniques are valid at various stages in the search process.Therefore,Mallipeddi et al.proposed ensemble of constraint handling techniques(ECHT)to solve the optimization problem with constraints.In this paper,by improving the integration of constraint processing technology proposed by Mallipeddi,two new ensemble of constraint handling techniques are proposed,combined with artificial bee colony algorithm and differential evolution algorithm,respective-ly.The first algorithm is an artificial bee colony algorithm based on ensemble of constraint handing techniques(ECHTABC).A new constraint handing technique is added to Mallipeddi’s ECHT.These five constraint handing techniques are combined into two ensembles of constraint handing techniques,which are used in employed bees phase and onlooker bees phase in ABC algorithm.Performance of ECHTABC has been tested on 28 benchmark functions in CEC 2017 and four classical engineering design problems.Experimental results show that ECHTABC can effectively improve the accuracy of solu-tion,It is more competitive than other state-of-the-art constrained optimization algorithms.The second algorithm is a differential evolutionary algorithm based on new ensemble of constraint handing techniques(NECHTDE).At the stage of generating new individuals,the algorithm adopts three different mutation strategies.Different constraint handing techniques are used to select new individuals,and local search is introduced to enhance the local optimization ability of the algorithm.Numerical experiments are carried out on 28 benchmark functions in CEC 2017 and the new algorithm is compared with several other advanced algo-rithms.The experiment results show that the NECHTDE algorithm performs better in solution accuracy.更多还原