群体智能优化算法—粒子群算法的研究和改进

发布时间：2018-06-25 08:29

本文选题：群智能优化算法 + 粒子群优化算法　；参考：《江南大学》2017年硕士论文

【摘要】：伴随着社会的不断进步,在经济管理、工程实践、科学研究等诸多领域出现了越来越多的优化问题,并且问题的复杂性、繁琐程度也是随之呈现迅猛的增长趋势。面对当今时代优化问题的非线性、大规模、高实时、低损耗等特点,传统的优化方法因为在方法设计之初面临着和当前不同的应用场景、计算机软硬件发展阶段不同、方法自身局限等各种因素,已难以在现代优化问题上取得很好的效果。通过大量学者专家的深入研究,群智能优化算法在求解现代优化问题时已展现出其优化速度快、精度高的特性,在工程实践中发挥着重要的作用。粒子群优化(Particle Swarm Optimization,PSO)算法是群智能优化算法中一个优秀的分支,其算法灵感来源于模拟社会型群居生物的觅食行为来获取最优解,算法在搜索中对外部信息量基本没有要求,粒子在高维空间中移动的依据来源于适应度函数值,是一种自适应智能优化算法。作为一种不确定型、概率型优化算法,PSO算法在面对各类优化问题时对目标函数没有例如连续性、可导性等较强的限制,表现出较好的通用性,并且在优化过程中是采用分布并行搜索的策略,因此具有较高的寻优效率。同时,由于PSO算法自身的局限性导致其在理论研究和实际应用方面还需进一步研究和改进。本文从粒子群优化算法的思想、机理出发,提出了更高效的改进算法并将其应用于工程实践中。在PSO算法的研究方面,本文立足于PSO算法的原理,针对粒子群算法搜索后期易陷入局部极值的缺点,提出了一种基于核矩阵协同进化的震荡搜索粒子群优化算法(Shock search particle swarm optimization algorithm based on kernel matrix synergistic evolution,KMSESPSO)。新算法对粒子进行了局部与全局结合的震荡搜索,且当整个粒子种群陷入停滞状态时,利用核矩阵对进化组进行协同进化以扩大种群的多样性。实验结果表明,新算法有效提高了粒子的全局搜索能力,既避免了粒子种群易早熟收敛,又较好地提高了寻优精度、加快了收敛速度,且有一定的鲁棒性。在吸收借鉴KMSESPSO算法的特点和思路上提出了基于Ransac协同震荡搜索混沌粒子群优化算法(Synergistic Shock particle swarm optimization algorithm based on Ransac With Chaos,RCSSPSO)。算法利用混沌的遍历性,运用到PSO提升其局部搜索有效性。为了增强全局优化能力,算法基于Ransac(Random Sample Consensus)思想,对群体做非线性震荡搜索。当种群陷入重停滞对协同进化集进行非线性协同进化,提升算法的搜索精度和种群的多样性。资源受限的项目调度问题(Resource-Constrained Project Scheduling Problem,RCPSP)普遍存在于各行各业,在项目执行过程中所有任务颗粒始终受到时序约束和资源约束,只有同时满足这两个约束任务才会被执行,RCPSP已被证明是NP-hard问题,研究RCPSP对项目进程预测与调整具有重要意义。针对资源受限的项目调度问题,本文项目调度采用基于粒子的拓扑排序和串行调度生成方案。为了进一步验证算法的有效性,本文和具体调度项目进行对比检验,实验结果表明算法在求解RCPSP时获得了更短的项目周期以及最短工期的平均偏差率较低,有更好的精度和鲁棒性,具有理论价值和工程实践意义。
[Abstract]:With the continuous progress of the society, more and more optimization problems have appeared in many fields, such as economic management, engineering practice, scientific research and so on. And the complexity and complexity of the problems also present a rapid growth trend. Facing the nonlinear, large-scale, high real-time, low loss and so on, the traditional optimization is the face of the optimization problems of the times. The method is faced with different application scenes at the beginning of the method design, the development stage of computer hardware and software is different, the method itself is limited and so on. It is difficult to achieve good results on the modern optimization problem. Through the deep study of a large number of scholars and experts, the swarm intelligence optimization algorithm has been displayed in solving modern optimization problems. It plays an important role in engineering practice. The Particle Swarm Optimization (PSO) algorithm is an excellent branch of the swarm intelligence optimization algorithm. The algorithm is inspired by the foraging behavior of the social group living organisms to obtain the optimal solution, and the algorithm is in the search of the outside. The amount of information is not required. The basis of the movement of particles in high dimensional space is based on the fitness function value. It is an adaptive intelligent optimization algorithm. As an uncertain and probabilistic optimization algorithm, the PSO algorithm has not such strong limitations as continuity and conductance, such as continuity and conductance in the face of various optimization problems. In the process of optimization, the strategy of distributed parallel search is adopted in the optimization process, so it has high optimization efficiency. At the same time, because of the limitations of PSO algorithm itself, it needs further research and improvement in theory and practical application. This paper proposes a more efficient method from the idea and mechanism of particle swarm optimization algorithm. In the research of PSO algorithm, based on the principle of PSO algorithm, this paper proposes a particle swarm optimization (Shock search particle swarm optimization algorit) algorithm based on kernel matrix co evolution. HM based on kernel matrix synergistic evolution, KMSESPSO). The new algorithm conducts a local and global concussion search for particles, and when the whole particle population is in a stagnant state, the kernel matrix is used to co evolve the evolution group to expand the diversity of the population. The experimental results show that the new algorithm effectively improves the overall particle size. The search ability not only avoids the premature convergence of the particle population, but also improves the optimization precision, speeds up the convergence speed, and has certain robustness. In the absorption of the characteristics and ideas of the KMSESPSO algorithm, a Ransac cooperative concussion search chaotic particle swarm optimization (Synergistic Shock particle swarm optimization al) is proposed. Gorithm based on Ransac With Chaos, RCSSPSO). The algorithm uses chaos ergodicity to improve the local search effectiveness of PSO. In order to enhance the global optimization ability, the algorithm is based on the idea of Ransac (Random Sample Consensus) to do nonlinear oscillation search for the group. When the group is trapped in the heavy stagnation, the cooperative evolution set is nonlinear synergy. Resource-Constrained Project Scheduling Problem (RCPSP) is common in all walks of life. In the process of project execution, all task particles are always constrained by temporal constraints and resource constraints, and only the two constraint tasks will be met at the same time. Implementation, RCPSP has been proved to be a NP-hard problem. The study of RCPSP is of great significance to the prediction and adjustment of the project process. In this paper, the project scheduling problem is based on particle based topological sorting and serial scheduling for resource constrained project scheduling problem. In order to further verify the effectiveness of the algorithm, this paper compares the scheduling project with the specific scheduling project. The test results show that the algorithm has a shorter project cycle and a lower average deviation rate in the shortest work period, and has better accuracy and robustness, which has theoretical value and engineering practical significance when solving RCPSP.
【学位授予单位】：江南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【相似文献】