Enhancing Rolling Horizon Production Planning Through Stochastic Optimization Evaluated by Means of Simulation

Abstract

Production planning must account for uncertainty in a production system, arising from fluctuating demand forecasts and execution-level friction. This article integrates scenario-based stochastic programming into a rolling horizon framework for capacitated lot sizing, evaluated via discrete-event simulation. We compare this stochastic approach against deterministic optimization and standard Material Requirements Planning (MRP) across varying customer update behaviors, shop loads, and diverse multi-stage topologies (divergent, convergent, and mixed). To accurately capture shop-floor dynamics, the framework introduces a non-anticipativity parameter controlling schedule flexibility, alongside probabilistic setup-time feedback and soft overtime constraints. Results indicate that optimization consistently outperforms MRP. In unbuffered, highly congested settings, stochastic optimization natively smooths workloads and reduces costs by up to 68%. However, introducing explicit safety stocks fundamentally shifts system dynamics: physical buffers effectively absorb shop-floor noise, diminishing the stochastic model's anticipative advantage and enabling deterministic optimization to dominate. Ultimately, this study offers critical managerial insights for aligning planning algorithms, inventory buffering, and schedule flexibility.