在企业经营过程中作为“第三方利润源”的物流成本在经营成本中占据很大比重。随着世界经济全球化发展，现代企业的经营范围也逐步扩大，物流网络更加趋向于大规模化、全球性。合理的规划不仅能为企业节约大量的运营成本，还能提高企业的服务效率，进而带来顾客满意度的提升和市场占有率的提高。在物流网络范围扩大和需求点增多的背景下，网络规划工作更为困难。当前设施选址与物流网络规划领域针对全球性大规模物流网络的研究数量较少，相关理论与应用研究尚不成熟。 本文以深圳某企业全球物流网络规划为背景，研究了面向全球配送的设施选址及多层级物流网络规划问题。该问题在供应商、配送中心、需求点三层基础上将全球化条件下多路径选择、多种运输方式结合的分段路径特点引入网络规划作业中，形成更为复杂的多层级网络结构。在计算成本时考虑了物流配送网络中的仓储、运输、人力等因素，综合衡量配送中心成本和物流网络整体运营成本。规划结果需要在候选城市中选取限制个数以内的城市设立配送中心，并且将所有需求点分配，同时要在有供求关系的节点之间多条路径中选取一条，使得系统的总体运营成本最小化。 本文根据全球物流配送网络的成本构成与限制条件设计了非线性整数规划数学模型，综合了设施选址、需求点分配与路径选择问题。由于该模型为NP-hard问题（Non-Deterministic Polynomial Problems），对于小规模数据本文利用CPLEX数学优化软件求解，同时验证了模型的有效性。在大规模数据的求解方面改进了模拟退火算法，首先处理数据得到每个需求点的可分配城市，然后在降温过程中利用需求点分配关系的随机更换产生邻接解，当降温达到一定次数或者最优解未更新累计达到一定次数时算法结束。 通过随机值算例对CPLEX与模拟退火算法求解的准确性与效率进行了对比，结果显示模拟退火算法能够获得近似最优解，在求解时间和求解规模方面具有较大的优势。同时本文选取了东亚和东南亚8个国家的部分城市构建仿真网络，利用模拟退火算法实现了问题的求解，并将相关结果直观的展示在百度地图上。
Logistics cost which is called the third profit source occupies a large proportion of the operating cost in a company. As the development of globalization, modern companies expand market and their logistics networks are getting larger and more global. Rational logistics network planning can not only save a lot of operating cost for a company, but also improve the service efficiency. As a result the customer satisfaction degree and the market occupancy will be improved. In the background of larger size logistics network and more demand points, the network design is also getting more difficult. In the area of location assignment and logistics network planning there are few studies about global distribution network. The theories and application researches are also not mature. Based on the global network planning of a company in Shenzhen, this paper studied the problem of facility location assignment and multi-level logistics network planning for global distribution. In addition to the suppliers, distribution centers and demand points, this paper also takes multi-route choice and segmented routes with different transportation types into consideration to design the network which has more complex structure. We considered the warehouse cost, the transportation cost and the human cost to calculate the distribution center cost and integrated operating cost. The targets are to select limited amount cities to establish distribution centers, assign the demand points to the distribution centers and choose one route between two connected nodes to minimize the whole operating cost. We designed a nonlinear integer programming model according to the relative costs and restrictions. The problem is NP-hard（Non-Deterministic Polynomial Problems）so we use the CPLEX Optimizer to get the exact optimal solution of the small scale data and verify the validity of the model. To solve the larger size problem we improved the simulated annealing algorithm. Firstly we get the assignable cities for each demand points from the data. Then generate the adjacent solution by randomly change the assignment of one demand point during the cooling process. The algorithm stops when the cooling times reach a certain number or the optimal solution do not changes for a certain times. Using the random value examples we tested the veracity and efficiency of CPLEX and simulated annealing algorithm. The result showed that simulated annealing algorithm could get an approximate optimal solution and had great advantage in solving time and scale. We also used some cities from eight countries in East Asia and Southeast Asia to build a simulation network. The problem is solved with simulated annealing algorithm and the result is shown on the Baidu Map.
