用基于族群的方法求解动态优化问题

发布时间：2018-04-18 13:54

本文选题：动态优化 + 族群方法　；参考：《中国科学技术大学》2017年硕士论文

【摘要】：与一般的优化问题相比,动态优化问题的特点是问题的状态(目标函数、约束条件)随时间变化。为了能够快速地捕捉到环境的变化,算法需要持续地定位和追踪最优解的移动。演化算法因其具有群体搜索的特点,适合求解一些复杂的问题,比如动态优化问题。基于族群的方法是一种有效的演化计算技术,已成为动态优化研究领域的热点之一。基于族群方法的基本思想是,将种群划分为若干个族群,不同族群在搜索空间的不同区域同时搜索。由于该方法允许种群同时定位多个最优解,因此更容易实现对全局最优解的追踪。本文主要研究使用基于族群的方法求解动态优化问题,研究内容主要包括如下两个方面。(1)提出了一个基于族群与记忆集的混合粒子群优化算法。该算法的特点是:用于更新种群的记忆个体的数量与族群数量相关并且随族群数量自适应地变化;限制每个族群被替换的个体数量不超过1;对提取的记忆个体分类处理,目的是在改善已有族群搜索能力的同时加强种群对潜在最优区域的搜索。在MPB、CMPB、DRPBG基准问题上对该算法测试并与其它算法进行比较,实验结果表明该算法是一个有竞争力的动态优化算法。此外,实验部分还讨论了记忆集的大小对结果的影响。(2)提出了一个应用于动态优化的族群划分方法psfNBC。与基本的Nearest-Better Clustering(NBC)算法相比,该算法的特点是:识别族群种子的过程只涉及部分个体而不是整个种群;种群按照最近种子的原则重新划分;缩放因子φ使用随机值而不是固定值。在识别族群种子时,本文提出了两种确定离群点数量的方法,即固定地和自适应地。此外,本文还给出了一个基于族群的粒子群算法框架,使用该框架对psfNBC以及其它几个有代表性的族群划分方法在MPB问题上测试,结果表明psfNBC可以在大多数的测试实例中取得最好的结果。
[Abstract]:Compared with the general optimization problem, the characteristics of dynamic optimization is the problem of the state (the objective function, constraint condition) change with time. In order to be able to quickly capture the changes in the environment, the algorithm needs to be continuous positioning and tracking the optimal solution of the mobile. Evolutionary algorithms because of its characteristic of population search, suitable for solving some complex the problems, such as dynamic optimization problems. The method is based on the group computing technology is an effective evolution, it has become a hot research topic in the area of dynamic optimization. The basic idea is based on the method of population, the population is divided into several groups of different ethnic groups in different regions of the search space and search. Because this method allows the population at the same time localization of multiple optimal solutions, it is easier to achieve the global optimal solution of the track. This paper mainly studies the use of dynamic optimization method for solving the problem of ethnic group based on the main research contents To include the following two aspects. (1) proposed a hybrid particle swarm optimization algorithm based on ethnicity and memory. The characteristic of this algorithm is used to update the number: individual and collective memory of population and population related changes with the number of individuals adaptively; each group was replaced by a limit of not more than 1; to classify the extracted individual memory, at the same time to improve the existing search ability in ethnic populations to strengthen the potential optimal searching area. In MPB, CMPB, DRPBG benchmark problems of the algorithm are tested and compared with other algorithms. The experimental results show that the algorithm is a competitive dynamic optimization algorithm. In addition, the experimental part of the memory set size on the results is also discussed. (2) proposed a psfNBC. group classification method is applied to the dynamic optimization and the basic Nearest-Better Clustering (NBC) algorithm. Than, the characteristic of this algorithm is: the process of identifying the seed groups involving only a part of the individual rather than the entire population; population according to the principle of seed recently re division; Phi zoom factor using random values rather than a fixed value. In recognition of ethnic seed, this paper puts forward two kinds of methods to determine from the group number, namely fixed and adaptive. In addition, this paper also gives a group based on particle swarm algorithm framework, using the framework of ethnic division method of representative test on MPB of psfNBC and several other, the results show that psfNBC can achieve the best results in most test cases.

【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【相似文献】