面向大规模库存路径优化问题的多信息协同的进化算法研究

Inventory routing problem (IRP) is the integrated optimization of inventory and transportation to reduce cost and increase benefits on the logistics management. IRP is an NP hard problem, and the trade-off relation between inventory and transportation makes the IRP more complex, especially the large-scale IRP which is closed to the actual demand is difficult to be optimized. To solve the problem, this project firstly chooses the Changhe Jiujiang factory as the research object, and constructs the large scale IRP optimization model with the different characteristics such as many products, many periods, random demand on the basis of the theoretical analysis and investigation. The large scale IRP is based on mixed integer nonlinear programming which can be solved by evolutionary algorithms. Then, the project does research on the evolutionary algorithm which is based on multi-information cooperation. The feedback of multi-information can reflect the state of the process of evolutionary computation more fully and comprehensively, and the cooperation of differentiation can solve the problems and bottlenecks arising from the calculation process more easily. Finally, the project employs parallel computing based on Spark cloud platform as a means of acceleration and the evolutionary algorithm on the basis of multi-information cooperation as an optimization tool for the purpose of solving the problem of complex large scale IRP. The project not only will enrich the evolutionary algorithm and the IRP studies, but also promote the development of modern logistics industry, and has important scientific significance and application value.

库存路径问题（IRP）是对库存和运输进行整合优化以实现物流管理降本增效的一类NP难问题。库存和运输的“效益背反”关系使得IRP变得愈加复杂，特别是贴近实际需求的大规模IRP求解难度更大。针对这一问题，本项目以昌河汽车九江工厂为蓝本，首先在现场调研论证与理论分析的基础上，基于混合整数非线性规划，构建适用于进化算法求解的具有多产品、多周期和随机需求等特征的大规模IRP模型。然后，研究多信息协同进化算法，多元信息的反馈可以更充分和全面地反映进化计算的过程状态，差异化的协同方式更易于解决计算过程中出现的问题和瓶颈。最后，基于Spark云平台实现并行加速，研究以高效的多信息协同进化算法对大规模IRP问题进行求解且具有时效性的可行方案，从而实现降低企业供应链成本的目标。本项目的研究，不仅可以进一步丰富进化算法和IRP研究的内涵，也将推动现代物流相关产业的发展，具有重要的科学意义和应用价值。

库存路径问题（IRP）是对库存和运输进行整合优化以实现物流管理降本增效的一类NP难问题。项目组根据研究计划，面向大规模库存路径优化问题对多信息协同的进化算法开展了相关研究工作，主要研究内容包括：.（1）基于混合整数非线性规划，构建了适用于进化算法求解的大规模复杂IRP模型，使用基于交付路线的解的表示方法来解决复杂的一对多IRP问题，设计了适合于IRP求解的基于差分进化算法的方法，在可行检验方法、局部搜索方法和最优路径搜索方法等方面进行了探索；.（2）研究了多信息协同进化算法，依据目标函数值、种群分布情况、种群更新比率以及变量相关性等进化环境信息，改单一的反馈为多信息的反馈，设计多信息协同的新型进化机制，并应用到了差分进化算法、人工蜂群算法、布谷鸟搜索算法、头脑风暴优化算法以及萤火虫算法等。.（3）研究了基于Spark云平台实现并行加速方法，设计了基于Spark的分布式并行差分进化算法、布谷鸟搜索算法和头脑风暴优化算法，提出了以分布式头脑风暴优化算法对大规模IRP问题进行求解且具有时效性的可行方案。.本项目的实施有助于推动进化算法的理论研究与应用实践的发展。