Abstract: To solve single product and multi-level problem of high transportation cost caused by stochastic demand under original equipment manufacturer, allowing opening or closing plants and warehouses, a new model is developed with the capacity and minimum order quantity constraints of plants, the capacity of warehouses and trucks. To minimize the whole supply chain cost, a Modified Genetic Algorithm（MGA） is constructed to solve the problem, while lingo is used to find the optimal solution. The two algorithms are compared by changing the number of plants, warehouses and customers. MGA is also compared with Genetic Algorithm（GA）. Numerical experiment results show that the gaps between MGA and lingo are below 10%, but the time is apparently less than lingo. MGA is better than GA both in cost and time.

Key words: stochastic demand, opening or closing, re-allocation, modified genetic algorithm, total cost

摘要： 针对单产品多层次生产外包环境下企业面对顾客需求随机导致运输成本过高的问题，采取允许代工厂和仓库可开关以及物流再分配方法，建立考虑代工厂的能力和最小加工批量、仓库转运能力以及货车运输能力的决策模型，设计加入STP（Shortest Travel Path）快速搜索策略的MGA（Modified Genetic Algorithm）进行求解。在变化工厂总数、仓库总数、顾客总数的不同维度下，将MGA与GA和LINGO分别进行对比。分析表明，MGA与LINGO求得的最优解之间差距均在10%以内，但求解时间明显优于LINGO求解时间，MGA在求解结果和时间方面均优于GA。

关键词: 需求随机, 可开关, 物流再分配, 改进型遗传算法, 总成本