Capacity Expansion under Forecast Uncertainty: Stochastic Integer Programming Approaches

该项目的目的是在预测的规划数据不可靠的情况下发展规划能力扩大的最佳技术。 随机规划已成为解决数据不确定性规划问题的重要工具。 然而，在容量扩张问题中，战略决策的整体性阻止了标准分解方法的使用，这些方法在随机线性规划中已经取得了成功。 这个项目将为容量扩展应用中出现的随机整数规划开发有效的解决方案。 该项目的一个关键组成部分将是确定可在解决战略中加以利用的特殊问题结构。 结构的结果将用于设计，分析和实施近似和精确的解决方案的算法。开发的方法的可行性将在重要的经济部门，如半导体晶圆制造设施和网络托管企业中得到证明。扩大产能以满足预期的需求增长是所有工业部门的一个关键战略问题。在高增长-高波动性行业，如IT行业，成本、需求和技术演变预测的不确定性以及扩张成本的规模经济使得产能扩张决策非常复杂。 使用随机整数规划的概念，本研究项目将开发一个基于优化的范例，以帮助明确解决预测的不确定性的能力扩展。 如果成功，该项目将提供强大的计算技术，以帮助各种行业的战略能力规划。预计从本研究中获得的见解将大大推进当前解决多阶段随机整数规划的最新技术。

The project is aimed at the development of optimization techniques for planning capacity expansions when forecasted planning data are unreliable. Stochastic programming has emerged as an important tool for solving planning problems with data uncertainties. In capacity expansion problems, however, the integral nature of strategic decisions prevents the use of standard decomposition approaches that have been successful for stochastic linear programs. This project will develop efficient solution strategies for stochastic integer programs arising in capacity expansion applications. A key component of the project will be to identify special problem structures that can be exploited within solution strategies. The structural results will be used to design, analyze, and implement approximate and exact solution algorithms. The viability of the developed methodology will be demonstrated in important economic sectors such, as semiconductor wafer fabrication facilities and web hosting enterprises.Capacity expansion to meet anticipated demand growth is a key strategic concern in all industrial sectors. In high growth-high volatility industries, such as the IT sector, uncertainties in forecasts for costs, demands, and technology evolution, and the economies-of-scale in expansion costs make capacity expansion decisions very complex. Using stochastic integer programming concepts, this research project will develop an optimization based paradigm for aiding capacity expansion that explicitly address forecast uncertainty. If successful, the project will provide robust computational techniques to aid strategic capacity planning in a wide variety of industries. It is also anticipated that insights gained from this research will significantly advance the current state-of-the-art in solving multi-stage stochastic integer programs.