Systematic Management of Uncertainties in Process Operations

计划、调度和供应链管理是过程工业中关键的运营活动。它们在降低成本和库存以及增加利润和响应能力方面发挥着重要作用。在现实中，不确定性普遍存在于工艺操作问题中。它的外观范围从制造过程到动态市场。成功的流程操作在很大程度上依赖于处理不确定性的能力，因为在确定性假设下做出的决策对于实际实现来说可能是次优的，甚至是不可行的。然而，解决过程操作中的不确定性仍然面临以下主要挑战：1)即使是中等数量的不确定参数也会导致超过当前计算能力的大量不确定情景；2)不确定参数之间经常存在相关性，这需要新的数学建模和求解技术来进行分析和优化；3)许多工艺操作问题涉及组合和非线性优化，在不确定条件下难以求解。**为了应对上述挑战，将探索一个管理工艺操作不确定性的系统框架。拟议的研究将包含若干创新要素。首先，我们将研究各种不确定性分析技术来量化不确定性的影响。其次，将开发一种通用场景约简算法，以选择不确定数据的代表性实现并考虑整体不确定性分布。第三，我们将开发新的鲁棒和随机优化算法，能够以计算高效的方式严格解决决策中的独立和相关不确定性。最后，提出的研究方案将生成一套扩展的优化工具，用于不确定条件下工艺运行的综合分析和优化

Planning, scheduling and supply chain management represent the key operational activities in the process industry. They play a major role in reducing costs and inventories, as well as increasing profits and responsiveness. In reality, uncertainty widely exists in those process operations problems. Its appearance ranges from the manufacturing process to the dynamic market. Successful process operations rely heavily on the ability to handle uncertainty since decisions made under deterministic assumptions can be suboptimal or even infeasible for practical implementations. However, addressing uncertainty in process operations still faces the following major challenges: 1) even a moderate number of uncertain parameters can lead to a huge number of uncertainty scenarios that exceed current computational capabilities; 2) correlations between uncertain parameters often exist, which require novel mathematical modeling and solution techniques for analysis and optimization; 3) many process operations problems involve combinatorial and nonlinear optimization, and it is difficult to solve such problems under uncertainty.**To address the above challenges, a systematic framework for managing uncertainties in process operations will be explored. The proposed research will contain several innovative elements. First, we will investigate various uncertainty analysis techniques to quantify the impact of uncertainties. Second, a generic scenario reduction algorithm will be developed to select representative realizations of uncertain data and to account for the overall uncertainty distribution. Third, we will develop novel robust and stochastic optimization algorithms that can rigorously address both independent and correlated uncertainties for decision making in a computationally efficient manner. Finally, the proposed research program will generate a set of extended optimization tools for comprehensive analysis and optimization of process operations under uncertainty.**