A BILEVEL DECOMPOSITION ALGORITHM FOR LONG-RANGE PLANNING OF PROCESS NETWORKS

The solution of the multiperiod MILP model for long-range planning of process networks by Sahinidis et al. (Comput. Chem. Eng. 1989, 13, 104 9) is addressed in this paper. The model determines the optimal select ion and expansion of processes over a long-range planning horizon, inc orporating multiple scenarios for varying forecasts for demands and pr ices of chemicals. A rigorous bilevel decomposition algorithm is propo sed to reduce the computational cost in the multiperiod MILP model. Th e decomposition algorithm solves a master problem in the reduced space of binary variables to determine a selection of processes and an uppe r bound to the net present value. A planning model is then solved for the selected processes to determine the expansion policy and a lower b ound to the objective function. Numerical examples are presented to il lustrate the performance of the algorithm and to compare it with a ful l-space branch and bound method.