Abstract

Scheduling multiproduct pipelines is a complex managerial task with a remarkable impact on the total revenues of the pipeline industry. It consists of sequencing, sizing and timing of injections and removals, to meet product demands on time at minimum cost. Generating the detailed scheduling of pipeline networks can be very challenging, requiring efficient optimization tools to find good feasible solutions. This paper develops a hierarchical decomposition approach for tree-like pipeline systems with two-level branching. Decisions related to the sequence of product injections and the destination for each batch, are made at the higher planning layer, while the lower layer then finds the sequence and timing of product deliveries. Each layer is tackled by a mixed-integer linear programming (MILP) formulation, which neither discretizes the time horizon nor divides a pipeline segment into packs of equal size. Solutions to three case studies present significant reductions in both the operating cost and the computational burden.