差分混合蛙跳算法的改进及其应用

发布时间：2018-01-03 22:26

本文关键词：差分混合蛙跳算法的改进及其应用　出处：《广东工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：计算智能优化算法是对自然界智慧和人类智慧的模仿,因其智能性、并行性和健壮性,具有很好的自适应能力和很强的全局搜索能力,得到众多研究者的广泛关注.混合蛙跳算法(SFLA)是一种新兴的智能优化算法,该算法结合了模因算法的局部启发式搜索和粒子群优化算法的全局搜索两者的优点,在进化过程中先进行局部精确搜索,再利用子群个体间的信息共享进行全局搜索,两者相互结合直至找出全局最优解.SFLA结构简单容易理解、控制参数少,具有很强的全局搜索能力.差分进化算法也是一种新兴的全局优化算法,局部更新策略类似于遗传算法,采用差分变异操作、交叉操作和选择操作更新产生新个体.经过一代代反复不断的局部进化,算法的搜索方向慢慢向全局最优解的方向靠近.DE算法具有精确的局部搜索能力,鲁棒性较强,已成为智能优化算法的重要分支.目前,将差分进化算法的局部更新策略与其他优化技术相结合来提高算法的优化性能,已被广泛应用于各个领域,在科学研究和生产实践中发挥着重要的作用.本文针对混合蛙跳算法在寻优过程中易陷入局部最优和早熟收敛的缺点,利用差分进化算法的局部精确搜索的特点和蛙跳算法强大的全局搜索能力融合提出一种改进的差分蛙跳算法(DSFLA).该算法借鉴差分进化中的变异交叉思想,在前期利用子群中其他个体的有用信息来更新最差个体,增加局部扰动性,以提高种群的多样性;在后期为加快收敛速度使用最好个体的信息进行变异交叉操作.同时在每一次产生新个体后,都要进行改进的越界处理来动态调整变化尺度,再与子群最差个体进行选择操作选出适应值更优的个体.本文还使用归档集进一步保留种群的多样性.通过对五个典型的连续优化函数进行实验仿真,测试结果表明DSFLA无论是在求最优解的稳定性上还是质量上都明显胜于SFLA和SFLA-AV,在前期保持种群多样性和后期提高收敛速度避免算法早熟都起到了较好的效果.最后,本文将改进的DSFLA运用在物流中求解带容量约束的车辆路径优化问题上,采用实数编码方式初始化种群,利用DEB规则处理约束问题,实验仿真得到多种优化路径,可为实际物流问题提供多种调度方案.
[Abstract]:Computational intelligence optimization algorithm is an imitation of natural intelligence and human intelligence, because of its intelligence, parallelism and robustness, it has good adaptive ability and strong global search ability. The hybrid leapfrog algorithm (SFLAs) is a new intelligent optimization algorithm. This algorithm combines the advantages of local heuristic search of meme algorithm and global search of particle swarm optimization algorithm. Then the information sharing among subgroups is used for global search, and the two are combined to find the global optimal solution. SFLA structure is simple and easy to understand, and the control parameters are less. Differential evolution algorithm is also a new global optimization algorithm, local update strategy is similar to genetic algorithm, using differential mutation operation. Crossover operations and selection operations update to produce new individuals. After generations of repeated local evolution, the search direction of the algorithm slowly to the direction of the global optimal solution close to the .DE algorithm has an accurate local search ability. Because of its strong robustness, it has become an important branch of intelligent optimization algorithm. At present, the local update strategy of differential evolution algorithm is combined with other optimization techniques to improve the optimization performance of the algorithm. It has been widely used in various fields and plays an important role in scientific research and production practice. This paper aims at the shortcomings of hybrid leapfrog algorithm which is prone to fall into local optimum and premature convergence in the process of optimization. Using the characteristic of local exact search of differential evolution algorithm and the powerful global search ability fusion of leapfrog algorithm, an improved differential leapfrog algorithm (DSFLAs) is proposed. The algorithm draws lessons from the idea of mutation crossover in differential evolution. In the early stage, the useful information of other individuals in the subgroup can be used to update the worst individual and increase the local disturbance, so as to improve the diversity of the population. In order to speed up the convergence of the best individual information mutation crossover operation. At the same time after each generation of new individuals must be improved cross-border processing to dynamically adjust the scale of change. Then the selection operation with the worst individuals of the subgroup is carried out to select the individuals with better fitness. The archival set is also used to further preserve the diversity of the population. Five typical continuous optimization functions are simulated experimentally. The test results show that DSFLA is superior to SFLA and SFLA-AV in the stability and quality of the optimal solution. Maintaining population diversity in the early stage and increasing convergence rate in the later stage to avoid premature algorithm have played a good effect. Finally. In this paper, the improved DSFLA is used to solve the vehicle routing optimization problem with capacity constraints in logistics, the real coding method is used to initialize the population, and the DEB rule is used to deal with the constraint problem. Simulation results show that many optimal paths can be obtained, which can provide a variety of scheduling schemes for practical logistics problems.
【学位授予单位】：广东工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【参考文献】