Procurement scheduling under supply and demand uncertainty: Case study for comparing classical, reactive, and proactive scheduling

Supply chain of a manufacturing system contains procurement activity, and unloading raw materials from delivery vessels to storage tanks should be scheduled optimally, subject to the operational constraints. In general, an MILP model is used for a systematic procurement scheduling. However if there exists significant uncertainty in supply and demand, the solution obtained from the deterministic model may be suboptimal or even infeasible. Therefore in this study, two alternative approaches are formulated to consider these uncertainties: reactive rescheduling in the rolling horizon manner, and Markov decision process (MDP) formulation based scheduling that incorporates future uncertainty into the scheduling directly. In order to solve the MDP problem, algorithmic approximation strategies (such as approximate dynamic programming) are studied and applied for reducing computational challenges. Finally, their performances are compared with those of the original MILP model for a simple case study.