Optimizing large-scale multiproduct pipeline scheduling: a novel hierarchical rolling-horizon framework

As a critical infrastructure in modern energy supply chains, the pipeline transportation enables efficient and safe long-distance delivery of liquid fuels. However, the optimal scheduling of multiproduct pipelines presents formidable computational challenges due to the combinatorial explosion of binary decision variables and rigorous operational constraints. To address this challenge, this paper develops an innovative hierarchical optimization framework that integrates (1) initialization with approximate modelling to quickly identify feasible solution space, (2) rolling-horizon optimization with hybrid discrete-time modelling to eliminate redundant binary variables and constraints, and (3) modification using continuous-time modelling to re-optimize the detailed schedules. Numerical experiments on five industrial-scale cases demonstrate that the proposed approach is able to obtain the optimal detailed schedules within 90 s, even where conventional full-scale MILP models fail to converge. It achieves a 71.3%–88.4% reduction in infeasible binary variables, leading to a computational time reduction of 88.2–99.3% compared to full-scale MILP models. Comprehensive parameter sensitivity analyses confirm the approach’s robustness and computational efficiency across varying parameter settings. This computational improvement makes it a practical decision-support tool capable of handling complex, large-scale and dynamic scheduling problems.