【作者】 胡静；

【导师】 谭瑛； 曾建潮；

【作者基本信息】 太原科技大学 ， 计算机应用技术， 2007， 硕士

【摘要】 在工业、社会、经济和管理等众多领域中,人们面临着大量的最优化问题。用模拟生物界自然现象而发展起来的群智能优化算法来解决此类问题已被越来越多研究者所关注。PSO(Particle Swarm Optimization)算法作为群智能算法的一个重要分支,由于算法简单易于实现已在许多领域得到了成功应用。PSO算法已成功地应用于各类静态函数的优化中。然而,真实世界遇到的问题往往是随时间变化的,频繁变化的解空间使得最好解随时间的变化而变化,当前时刻得到的最好解,不一定是下一时刻的最好解,这就需要对问题重新建模求解。所以,将微粒群算法应用到动态环境中,跟踪环境的变化并寻找不断变化的最好解具有积极且现实的意义。为了能跟踪到随环境变化而变化的最好解,动态环境下的PSO算法需解决两个问题:一是能检测到环境的变化,二是当环境变化后微粒能紧密地跟踪变化直到获得最好解,即响应环境的变化。本文从这两个方面对适应于变化环境的微粒群算法进行了详细的论述,主要的研究工作如下:1、提出了改进的环境检测方法——基于微粒自身信息的检测方法,不仅降低了的算法复杂度,而且弥补了常用环境检测方法无法准确检测出环境变化的局限性。2、提出了响应环境变化的响应依据——种群多样性和环境变化前后全局最好解的距离。分析了提出响应依据的原因、必要性及响应依据之间、响应依据和重设之间的关系。3、受标准PSO算法鸟类具有“社会认知”能力的启发,提出了响应环境变化的响应方法——基于学习的响应方法,该方法定义了环境变化后需要重设的部分微粒及其运动方向。最后,介绍了PSO算法在更复杂环境变化中的发展前景和主要研究方向。更多还原

【Abstract】 There are many optimization problems in the fields of industry, society, economy, management, and so on. As a new evolutionary technology, Swarm Intelligence Algorithm simulating some natural phenomenon, has made progresses on solving these optimization problems. As one of Swarm Intelligence Algorithm, Particle Swarm Optimization, with simple and programming easily, has already got the successful application in many realms.The PSO algorithm is already successfully applied in the optimization of various static functions. However, many real world problems are dynamic and stochastically change over time, the current reasonable optimum is not certainly the optimum in the next time. It is essential to re-design the model of problem over time. It is realistic and active meanings to track changes for dynamic environments and search the optimum after a change.In order to track the movement of optimum with changeable environments, the goal in dynamic environments is not only to detect the changes automatically but also to respond a variety of changes as timely as possible. From view of detection and response, a new improved and adaptive PSO has been introduced and discussed in detail in this paper. The main researches are as follows:1、An improved detection method at the particle level not only reduces the optimization cost but also makes up the limitation of the usual detection methods.2、Response conditions,including population diversity and Dgbest,are designed. The reasons as well as necessity of putting forward response conditions and relation with reset are analyzed.3、Motivated by birds“social cognition”from standard PSO model, a new response method, learning from the optimum for new environments, is designed. This method defines part of particles to be reset and their flying direction after a change.Finally, the developing and research trend of PSO in more complicated environments are pointed out.更多还原