Distributionally Robust Process Scheduling under Ambiguous Uncertainty

Abstract

Ahstract- Distributionally robust optimization (DRO) is paving a new way to decision-making under ambiguous uncertainties. However, it remains challenging to approach the process scheduling problem from a distributionally robust perspective. We propose a novel DRO scheduling model under demand uncertainties, and develop the solution strategy. A data-driven approach is first proposed to construct ambiguity sets. To account for the sequential decision-making structure in process operations, two-stage DRO problems are developed and affine decision rules are adopted to derive an approximate, albeit conservative solution. Applications in industrial-scale batch process scheduling demonstrate that, the proposed approach can effectively leverage data information, better hedge against the inexactness of uncertainty distributions, and bring more profits.