随着社区新零售的兴起，网格仓作为链接线上平台与线下门店的中转站，显得尤为重要。但网格仓配送环境复杂多变，送达时间难以保证，严重影响社区客户的线上购买需求，而社区线上需求分布的变动又会对网格仓的配送决策造成不可避免的影响，使其分拣及运输成本居高不下、难以盈利等问题突出。针对以上问题，本文在分析社区需求变动和网格仓配送决策的相互作用机制基础上，首先提出一种社区需求预测-网格仓配送决策迭代优化框架。其次构建基于支持向量回归算法的需求预测模型，实现社区订单需求量的预测。然后构建以利润最大化为目标的网格仓订单配送路径优化模型，并采用自适应大邻域搜索算法对模型求解。最后通过算例验证了迭代优化决策框架和相关模型方法的有效性和实用性。本文的研究成果将为社区新零售下网格仓的运营决策优化提供理论依据。

With the rise of new retail in the community， grid warehouse， as a transfer station linking online platforms and off line stores， is particularly important. However， online purchase demands of community customers are seriously affected by the changeable distribution environment of grid warehouses and the difficulty of delivery time. The decisions of grid warehouse operators are inevitably affected by the change of online demand distribution in communities， which brings high sorting and transportation costs and difficulties in profits making. Firstly， based on the analysis of interaction mechanism between the changes of community demand and the decision of grid bin distribution， a “community demand prediction and grid bin distribution decision making” iterative optimization framework is proposed in order to solve these problems. Secondly， a model of demand prediction based on support vector regression algorithm is constructed to predict the demand of community orders. Thirdly， an optimization model about distribution path of grid warehouse is constructed with the goal of profit maximization， using the adaptive large neighborhood search algorithm to solve it. Finally， an example with 35 communities is given to verify the effectiveness of the framework we proposed with the models and algorithms. Additionally， the changing rules of profit variation in different conditions is deeply explored based on the actual situation. It is shown that： （1） Using the iterative optimization framework considering different demands can increase total profit by about 50%. （2） In the stage of grid warehouse distribution decision， compared with the solution by Cplex solver， the Adaptive Large Neighborhood Search algorithm can Increase efficiency more than tenfold. （3） In the same region， the expansion of scale does not always lead to an increase in profits. As the number of pick-up points increases， the total profit shows a trend of “first increase and then decrease”. （4） For grid warehouse operators， it is recommended to use minivans and light trucks responsible for distribution tasks. The research achievement of this paper will provide a theoretical basis for the operation decision optimization of grid warehouse in the new retail environment.