中文摘要：

研究了一种货运站场的选址分配问题，它是一类容量受限且需求随机的多设施选址问题。针对该问题提出一种两阶段决策优化方法：第1阶段为选址决策，发生在站场实际建设前；第2阶段为分配决策，发生在站场建设后、实际运营前。选址和分配2个阶段在模型求解时同步完成。基于场景规划理论建立了对应的容量受限型数学模型，该模型的建设费用和运营费用同站场容量成比例，同时约束条件考虑了规划的阶段性需求，更贴近实际情况。设计了采用符号编码策略的遗传算法，算法的编码方式不仅极大缩小了搜索规模和计算复杂度，而且能消除很多遗传算法的实施限制。针对中国的一个国家级枢纽规划问题，应用两阶段优化方法，基于实际统计数据，建立具体的数学模型并运用所设计的算法进行求解，得到不同场景下的选址分配结果，给出了相应的情况分析。

英文摘要：

Freight station location-allocation problem was analyzed, which was a kind of capacitated multi-facility location problem （MFLP） with stochastic demands （SMCFLP）. A two-stage decision optimization approach was developed for solving this problem, in which the first stage dealt with the location decision that would be made before the construction of station, and in the second stage, the allocation decision was ideally made before actual operation but after station construction. Both stages are solved simultaneously by a single model. A corresponding mathematical model with capacity constraint was established based on scenario approach. The construction and operating costs in the model were assumed to be proportional to station capacity. Meanwhile, the constraints considered periodic demand of plan, closer to the actual situation. A genetic algorithm （GA） was developed using symbolic coding to search for the optimal solutions, where the coding method reduced the search scope and computational complexity and some restrictions of GA were eliminated. Aimed at one national level transportation hub planning problem in China, with two-stage optimization approach, based on actual statistical data, specific mathematical model is established and solved by the designed algorithm. Many kinds of optimization result of the location-allocation under different scenarios and related analyses are performed.