A new scheduling algorithm based on an optimal control theory is presented for large-scale industrial problems. The main feature of this idea lies in a two-level hierarchical structure consisting of a reference model and a rigorous model, respectively. As a fair compromise between the excess of decision detail and long-range planning inspiration, we address simplified or relaxed models to derive optimal trajectories. The optimal trajectory is then used as a reference trajectory to be tracked in real time by a feedback control system to cope with uncertainty and modeling errors. We apply this idea to operation scheduling of refinery processes that convert crude oil mix into a variety of marketable products through a number of refining processes. The optimal control formulation is shown to enhance solution performance compared with the traditional mixed-integer model.