人工鱼群智能优化算法的研究及应用

发布时间：2019-03-13 08:00

【摘要】：传统优化算法常常对最优化问题的解析性质有特定要求,难以通用地求解各类复杂多样的最优化问题。近年来的研究发现,群体智能优化算法在解决此类优化问题方面显示出特殊的优势,受到了许多学者的关注和重视。作为一种模仿鱼群觅食的智能优化算法,人工鱼群(Artificial Fish Swarm Algorithm,AFSA)算法不要求目标函数具备特有的解析性质,并对初值及参数值不敏感,具有并行处理能力以及良好的随机性等特点,目前已得到了广泛的研究和应用。本文针对人工鱼群算法进行了研究和改进,并将其应用在解决物流配送中心选址问题,具体内容概括如下:(1)人工鱼群算法存在易陷入局部极值、寻优精度不高等问题。针对这些问题,本文提出了反向自适应高斯变异的人工鱼群算法(Opposite Adaptive and Gauss Mutation Artificial Fish Swarm Algorithm,OAGMAFSA)。该算法通过引入反向解来调整人工鱼的移动方向及位置,提供更多的机会发掘潜在的较优空间,使人工鱼群能够快速跳出局部最优;同时为了更好地平衡全局搜索与局部搜索之间的关系,使用了一种非线性自适应视野步长策略;再者,为了增加鱼群的多样性,降低人工鱼陷入早熟的可能性,提出了一种最优解引导的高斯变异机制。仿真实验结果表明,该算法能有效地提高人工鱼群的寻优精度,并且避免了人工鱼群过早收敛。(2)研究发现,基本人工鱼群算法主要存在两方面不足:(1)人工鱼群算法无法根据鱼群在觅食过程中的鱼群分布情况自适应地控制视野和步长;(2)人工鱼群算法中的每个人工鱼的行为属于局部搜索,缺少全局性。为此,本文提出了精英学习的多维动态自适应人工鱼群算法(Elite Learning-based Multi-dimensional dynamic Adaptive Artificial Fish Swarm Algorithm,EMAAFSA)。算法通过为每个维度设定独立的视野和步长,从而定义了视野向量、步长矩阵及多维邻域,以此改进了鱼群的4种基本行为,使人工鱼个体能够根据鱼群分布情况自适应调整寻优范围。同时,提出了一种人工鱼精英学习策略增加了鱼群的全局性,降低了人工鱼陷入局部最优的可能性。仿真实验结果表明,该算法能有效地提高人工鱼群的寻优质量、鲁棒性,且提高了人工鱼群的全局搜索能力。(3)物流配送中心是负责贮藏各类货物,并且进行物品配送、装卸等工作的物品流通中心。多物流配送中心选址问题是带约束的非线性规划问题,本文将提出的人工鱼群优化算法应用于多物流配送中心选址问题中,并通过仿真实验得到了较好的结果,证明了OAGMAFSA和EMAAFSA算法有较高的有效性和应用价值。
[Abstract]:Traditional optimization algorithms often have specific requirements for the analytical properties of optimization problems, and it is difficult to solve all kinds of complex optimization problems in general. In recent years, it has been found that swarm intelligence optimization algorithm has special advantages in solving this kind of optimization problems, and has been paid attention to by many scholars. As an intelligent optimization algorithm, artificial fish swarm (Artificial Fish Swarm Algorithm,AFSA (artificial Fish Swarm) algorithm does not require special analytic properties of objective function, and is not sensitive to initial value and parameter value. With the characteristics of parallel processing ability and good randomness, it has been widely studied and applied. In this paper, the artificial fish swarm algorithm is studied and improved, and it is applied to solve the location problem of logistics distribution center. The concrete contents are summarized as follows: (1) the artificial fish swarm algorithm is easy to fall into the local extremum, and the optimization precision is not high. In order to solve these problems, an artificial fish swarm algorithm (Opposite Adaptive and Gauss Mutation Artificial Fish Swarm Algorithm,OAGMAFSA) based on reverse adaptive Gao Si mutation is proposed in this paper. In this algorithm, reverse solution is introduced to adjust the direction and position of artificial fish, which provides more opportunities to explore the potential optimal space, so that the artificial fish can jump out of the local optimum quickly. At the same time, in order to better balance the relationship between global search and local search, a nonlinear adaptive horizon step-size strategy is used. Furthermore, in order to increase the diversity of fish and reduce the possibility of artificial fish falling into early maturity, a mechanism of Gao Si variation guided by optimal solution was proposed. The simulation results show that the proposed algorithm can effectively improve the optimization accuracy of artificial fish and avoid premature convergence of artificial fish. (2) it is found that the proposed algorithm can effectively improve the optimization accuracy of artificial fish stocks and avoid premature convergence of artificial fish stocks. There are two main shortcomings in the basic artificial fish swarm algorithm: (1) the artificial fish swarm algorithm cannot control the field of view and the step size adaptively according to the fish colony distribution in the feeding process; (2) the behavior of each artificial fish in artificial fish swarm algorithm belongs to local search and lacks of global. Therefore, a multi-dimensional dynamic adaptive artificial fish swarm algorithm (Elite Learning-based Multi-dimensional dynamic Adaptive Artificial Fish Swarm Algorithm,EMAAFSA) for elite learning is proposed in this paper. By setting an independent field of view and step size for each dimension, the algorithm defines the visual field vector, step matrix and multi-dimensional neighborhood, thus improving the four basic behaviors of fish stocks. The artificial fish can adjust the optimal range adaptively according to the distribution of fish stocks. At the same time, an elite learning strategy for artificial fish is proposed, which increases the overall quality of fish and reduces the possibility of artificial fish falling into local optimization. The simulation results show that the algorithm can effectively improve the optimization quality, robustness and global search ability of artificial fish herd. (3) the logistics distribution center is responsible for storing all kinds of goods and delivering goods. An article circulation center for handling, etc. Multi-logistics distribution center location problem is a nonlinear programming problem with constraints. In this paper, the artificial fish swarm optimization algorithm is applied to multi-logistics distribution center location problem, and good results are obtained through simulation experiments. The validity and application value of OAGMAFSA and EMAAFSA algorithms are proved.
【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

【参考文献】