Comparison of Robust Optimization and Info-Gap Methods for Water Resource Management under Deep Uncertainty
Comparison of Robust Optimization and Info-Gap Methods for Water Resource Management under Deep Uncertainty
复制标题
DOI：
10.1061/(asce)wr.1943-5452.0000660
复制
发表时间：
2016-09
影响因子：
3.1
通讯作者：
T. Roach;Z. Kapelan;R. Ledbetter;Michelle Ledbetter
中科院分区：
文献类型：
--
作者：
T. Roach;Z. Kapelan;R. Ledbetter;Michelle Ledbetter
本文评价了两种已建立的决策方法，并分析了它们在水资源管理(WRM)问题中的表现和适用性。被评估的方法有信息缺口(IG)决策理论和稳健优化(RO)。选择这些方法主要是为了研究评估水系统对深度不确定性的稳健性的局部和全局对比方法，也是为了比较稳健性模型方法(IG)和稳健性算法方法(RO)，前者选择和分析一组预先指定的策略，后者使用优化算法自动生成和评估解决方案。该研究提出了一种新的基于区域的IG稳健性建模方法，并评估了未来流量气候变化预测在水资源适应规划情景生成中的适用性。这些方法被应用于一个类似英国苏塞克斯北部水资源区的案例研究，评估它们在改善基于风险的水资源管理问题方面的适用性，并突出每种方法在选择气候变化和未来需求不确定性下的适当适应战略方面的优势和劣势。这两种方法都产生了成本稳健性的帕累托集合，并强调RO为各种不同的目标稳健性水平产生了较低的成本策略。IG产生了更昂贵的帕累托策略，因为其更具选择性和更严格的稳健性分析，这是由于更复杂的场景排序过程造成的。
This paper evaluates two established decision-making methods and analyzes their performance and suitability within a water resources management (WRM) problem. The methods under assessment are info-gap (IG) decision theory and robust optimization (RO). The methods have been selected primarily to investigate a contrasting local versus global method of assessing water system robustness to deep uncertainty, but also to compare a robustness model approach (IG) with a robustness algorithm approach (RO), whereby the former selects and analyzes a set of prespecified strategies and the latter uses optimization algorithms to automatically generate and evaluate solutions. The study presents a novel area-based method for IG robustness modeling and assesses the applicability of utilizing the future flows climate change projections in scenario generation for water resource adaptation planning. The methods were applied to a case study resembling the Sussex North Water Resource Zone in England, assessing their applicability at improving a risk-based WRM problem and highlighting the strengths and weaknesses of each method at selecting suitable adaptation strategies under climate change and future demand uncertainties. Pareto sets of robustness to cost are produced for both methods and highlight RO as producing the lower cost strategies for the full range of varying target robustness levels. IG produced the more expensive Pareto strategies due to its more selective and stringent robustness analysis, resulting from the more complex scenario ordering process.