Grain drying capacity planning and scheduling under yield uncertainty: Minimizing post-harvest losses and operational costs

This paper investigates the joint optimization of grain drying capacity planning and scheduling under yield uncertainty, to minimize operational costs and reduce post-harvest losses. We propose a two-stage stochastic optimization model that integrates decisions on post-harvest service center (PHSC) selection, initial drying machine deployment, and grain origin allocation in the first stage. In the second stage, after yield realization, the model addresses machine supplementation and detailed drying task scheduling. To handle the complexity arising from scheduling feasibility constraints, we propose an exact solution approach based on logic-based Benders decomposition (LBBD). In this approach, scheduling feasibility is efficiently determined using the preemptive earliest due date (PEDD) rule, and customized Benders cuts are developed. Additionally, acceleration techniques, including multi-cut strategies, valid inequalities, and the minimal infeasible subset (MIS) method, are employed to enhance computational efficiency. We validate the effectiveness and scalability of the proposed solution approach through extensive computational experiments, which solve instances with up to 120 origins and 200 yield scenarios within one hour. Two sensitivity analyses are conducted to assess the impact of second-stage machine costs and different scheduling rules on resource utilization and total cost performance. A real-world case study based on data from Hubei Province in China further demonstrates the practical applicability of our model and method. In this case study, a data-driven approach is used to characterize yield uncertainty by fitting historical yield data to probability distributions to ensure realistic scenario generation.