Chemical Process Scheduling under Disjunctive Uncertainty Using Data-Driven Multistage Adaptive Robust Optimization

Process scheduling is one key layer of decision hierarchy for process industries to optimize their production schedule in order to gain the long-term economic viability. A main challenge of process scheduling lies in the treat of uncertainties when approaching the multistage adaptive robust optimization of the scheduling problem. In this work, we introduce the non-parametric Bayesian inference technique to construct the data-driven disjunctive uncertainty set to alleviate the over-conservatism issue faced by most commonly used fixed-shaped uncertainty sets, and utilized the piecewise linear decision rule to generate solutions for the multistage batch scheduling optimization. Based on improvement in uncertainty set construction and decision rule flexibility, we demonstrated with an industrial process case study that the proposed approach with the disjunctive uncertainty set and decision rule is capable of generating usually better process scheduling optimization solutions in comparison to those obtained by conventional adaptive robust optimization approaches.