Learning to Recycle Benders Cuts for Mixed Integer Model Predictive Control

Mixed integer MPC problems arise frequently in cases where the operation of a system depends on continuous and discrete decisions, leading to mixed integer optimization problems. However, the online solution of such problems is computationally challenging. In this work, we develop a machine learning approach to determine which cuts should be used as a warm start (recycled) for Generalized Benders Decomposition for the solution of mixed integer MPC problems. Computational results on a case study regarding the operation of chemical processes show that the proposed approach leads to a significant reduction in solution time (up to 40% reduction).