A Disjunctive Programming Model and a Rolling Horizon Algorithm for Optimal Multiperiod Capacity Expansion in a Multiproduct Batch Plant

Abstract: In this paper we address a multi-period mixed integer non-linear problem for the capacity expansion of multiproduct batch plants. In this problem, given a certain batch plant with its current configuration, product recipes, and growing production targets, modular expansions are wanted so that new demand can be met. Unlike most work for the batch retrofit problem found in literature, a multi-period disjunctive model is presented, so that long term investments and expansions can be planned out in advance. Although effective for short periods, the proposed model becomes computationally inefficient for long time horizons. To address this issue, we propose a rolling horizon algorithm that further exploits the advantages of a disjunctive programming model. A numerical example based on a case study from industry is presented that shows that the rolling horizon algorithm is very effective on finding near optimal solutions to large instances with a considerable number of time periods. Furthermore, empirical evidence shows how the solution found by the proposed algorithm can be used as a starting solution for the direct method for the original problem to deliver a global optimal solution to the problem.