Time Aggregation for Network Design to Meet Time-Constrained Demand

We study a network design problem inspired by a strategic planning problem encountered in the Hunter Valley Coal Chain. Demand is given in the form of freight that is available from a specific date and has to be transported from multiple origins to a single destination before its deadline. It is possible to temporarily store freight at certain intermediate locations along the way from origins to destination. The objective is to determine minimum-cost capacity expansions required on the links and nodes of the network, if any, so as to be able to transport all freight within its given time windows. A natural mixed integer programming formulation with a daily granularity quickly becomes computationally intractable. We investigate the potential of time aggregation to overcome the computational challenges. By aggregating consecutive time periods, a smaller instance is obtained, which can be solved more easily and provides a lower bound on the optimal value. A carefully designed iterative disaggregation scheme identifies a time aggregation that yields an optimal solution to the original problem. An extensive computational study demonstrates the efficacy of the proposed approach.