This paper studies a variant of the lot sizing problem that arises in the context of disaster management. In this problem, a fixed budget has to be allocated efficiently over multiple time periods to procure large quantities of a staple food that will be stored and later delivered to people affected by disaster strikes whose numbers are unknown in advance. Starting from the deterministic model where perfect information is assumed, different formulations to address the uncertainties are constructed: classical robust optimisation, risk-minimisation stochastic programming, and adjustable robust optimisation. Experiments conducted using data from West Java, Indonesia allow us to discuss the advantages and drawbacks of each method. Our methods constitute a toolbox to support decision makers with making procurement decisions and answering managerial questions such as which annual budget is fair and safe, or when storage peaks are likely to occur.

本文研究了灾害管理背景下出现的批量大小问题的一种变体。在这个问题中，必须在多个时间段内有效地分配固定预算，以采购大量主食，这些主食将被储存起来，然后运送给受灾害影响的人们，而这些人的数量事先是未知的。从假设完美信息的确定性模型开始，构建了解决不确定性的不同公式：经典鲁棒优化、风险最小化随机规划和可调整鲁棒优化。使用印度尼西亚西爪哇的数据进行的实验使我们能够讨论每种方法的优点和缺点。