Multiperiod MINLP models and methods for the optimal design and planning of process and supply networks

In this thesis, the optimal design and planning of process and supply networks is addressed through multiperiod Mixed Integer Nonlinear Programming (MINLP) models and methods. Multiperiod MINLPs become quickly intractable and this challenge is addressed with improved modeling as well as specialized solution algorithms.; A general disjunctive multiperiod nonlinear optimization model, which incorporates design, operation and expansion planning, is proposed. A disjunctive bilevel decomposition algorithm is developed for the solution of this model, and applications in the areas of general process networks and retrofit design of multiproduct batch plants are considered.; Furthermore, two multiperiod MINLP models for offshore hydrocarbon infrastructure design and planning are presented. For the solution of the first model, an iterative aggregation/disaggregation algorithm is proposed, together with a novel dynamic programming subproblem to improve the aggregation scheme at each iteration. In the second model, it is shown that the inclusion of complex fiscal rules leads to substantial increases in the Net Present Value (NPV) of the project, but also results in more than an order of magnitude increase in the computational effort required. This is addressed through a specialized heuristic algorithm that relies on Lagrangean decomposition.; An integrated solution methodology for production planning and reactive scheduling of a hydrogen supply network is proposed. Multiperiod MINLP models for both the planning and scheduling levels are presented, as well as a heuristic Lagrangean decomposition algorithm to deal with the computational challenges on the planning level.; Results show that in all cases the proposed algorithms lead to one or more orders of magnitude decrease in solution time compared to DICOPT++, and that previously unsolvable models are solved in reasonable time to yield good sub-optimal or optimal solutions.