REFORMULATION OF MULTIPERIOD MILP MODELS FOR PLANNING AND SCHEDULING OF CHEMICAL PROCESSES

A large number of planning and scheduling problems can be formulated as multiperiod MILP models which often require substantial computational expense for their solution. This paper presents and demonstrates the value of nonstandard formulations of such problems. Based on a variable disaggregation technique which exploits lot sizing substructures, we propose a strategy for the reformulation of conventional multiperiod MILP models. The suggested formulations involve more constraints and variables but they exhibit tighter linear programming relaxations than standard approaches. The proposed reformulation strategy is applied to a model for batch scheduling and a model for long-range planning. Numerical results are presented for these problems to demonstrate that-due to their tighter linear programming relaxations-the reformulations can lead to up to an order of magnitude faster computational results and make possible the solution of larger problems.