Short-term supply chain management in upstream natural gas systems

[摘要] (cont.) The resulting mathematical program is a mixed-integer nonlinear program (MINLP) with nonconvex functions and can be solved with the current state-of-the-art global optimization approaches, provided careful attention is paid to the model formulation.A hierarchical multi-objective approach is proposed to address multiple objectives when operating upstream systems, by optimizing a lower priority objective over the multiple optimal solutions of a program with a higher priority objective to obtain a win-win scenario. A reproducible case study that captures all the features of natural gas upstream systems is constructed to facilitate future work in algorithm development for such problems. A preliminary comparison with the existing approach indicates that substantial benefits may be possible by using the proposed approach for short-term planning. The application of a reduced-space global optimization approach to planning in upstream gas networks has also been demonstrated, which can significantly lower the number of variables in the branch-and-bound algorithm. The lower bounding problem is implemented using McCormick (convex) relaxations of computer evaluated functions and solved by implementing a nonsmooth bundle solver as a linearization tool to obtain a linear programming relaxation. The upper bounding problem is implemented using automatic differentiation and a local NLP solver. Branch-and-bound with reduction heuristics and linearization propagation is used for global optimization.This approach has been found to be competitive with current state-of-the-art global optimization algorithms for upstream planning problems.

[发布机构] Massachusetts Institute of Technology