集装箱班轮轴辐式网络区间规划模型研究

发布时间：2018-01-29 13:03

本文关键词： 区间规划不确定性混合遗传算法轴辐式网络　出处：《大连海事大学》2014年硕士论文　论文类型：学位论文

【摘要】：在集装箱轴辐式班轮网络设计中,一个重要参数是各港口间的集装箱运输需求。当需求发生变化时,优化设计的网络随之发生变化。优化设计的最终航线网络结构一旦决策就很难改变,以确定需求为模型构建参数的航线网络优化不能随市场运输需求的变化,使决策者承担相当大的风险。因此,需要考虑不确定性条件下的轴辐式网络优化。对于不确定班轮问题的处理,主要有随机规划、模糊规划、情景集鲁棒优化等方法。其中,随机规划方法需要不确定变量的概率分布,模糊规划需要隶属度函数、情景集法需要情景概率分布。港口的复杂性使实际的不确定变量分布难以获取。因此,以上方法有其相应的局限性。在此条件下,本文尝试引入区间集合形式约束集装箱运输需求参数,以运输成本和中转成本总成本为目标函数,建立混合整数线性区间规划,联合优化枢纽港选址、支线港配置、干线航线三个问题。引入风险因子,将含有区间形式的目标函数转化确定性函数,从而进行优化求解。本文建立的模型为NP-hard问题。模型的复杂性决定了问题求解的难度,鉴于此,提出利用GA(遗传算法)和AC(蚁群算法)相结合的混合遗传算法求解。其中,枢纽港选址利用遗传算法,支线港配置利用最短路径法,干线航线优化采用蚁群算法。但整个算法仍以遗传算法为框架,选取目标函数值的倒数作为适应度函数；选取最优个体和轮盘赌相结合的算子作为选择算子；采用单点交叉作为交叉算子；采用基因互换作为变异算子。最后利用算例验证模型的可行性和算法的有效性。同时也说明利用区间规划处理需求不确定性问题,不仅能体现数据非精确性,同时求解结果能包含不确定性信息并且在一定程度上反应集装箱需求信息,因此能够使决策者更为详细地了解风险状态与后果。
[Abstract]:In the design of container axis-spoke liner network, an important parameter is the container transportation demand between ports. The final route network structure of optimization design is difficult to change once the decision is made, and the route network optimization based on the model construction parameters can not change with the market transportation demand. Therefore, it is necessary to consider the radial network optimization under uncertain conditions. For the treatment of uncertain liner problems, there are mainly stochastic programming, fuzzy programming, scenario set robust optimization and other methods, in which the stochastic programming method needs the probability distribution of uncertain variables. Fuzzy programming needs membership function, scenario set method needs scenario probability distribution, and the complexity of port makes it difficult to obtain the actual uncertain variable distribution. Therefore, the above method has its corresponding limitations. Under this condition, this paper attempts to introduce interval set form to constrain container transportation demand parameters, take the total cost of transportation and transit cost as the objective function, and establish mixed integer linear interval programming. By introducing risk factors, the objective function with interval form can be transformed into deterministic function, which can be solved optimally by jointly optimizing the location of hub port, the configuration of branch port and the route of trunk line. The model established in this paper is a NP-hard problem. The complexity of the model determines the difficulty of solving the problem. A hybrid genetic algorithm based on GA (genetic algorithm) and AC (Ant Colony algorithm) is proposed, in which the location of hub port is based on genetic algorithm, and the shortest path method is used in the configuration of branch port. Ant colony algorithm is used in trunk route optimization, but the whole algorithm is still based on genetic algorithm, and the reciprocal of objective function is selected as fitness function. The operator which combines the optimal individual and roulette is chosen as the selection operator. Single point crossover is used as crossover operator. Gene exchange was used as mutation operator. Finally, the feasibility of the model and the validity of the algorithm are verified by an example. At the same time, the use of interval planning to deal with demand uncertainty can not only reflect the inaccuracy of the data. At the same time, the solution results can contain uncertainty information and reflect container demand information to a certain extent, so that the decision makers can understand the risk state and consequences in more detail.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U695.22

【参考文献】