Decomposition strategies for the optimal design of integrated sites and multi-site planning and scheduling

This dissertation proposes decomposition strategies to solve large-scale Mixed-integer Linear Programming (MILP) problems occurring in process networks made up of multiple production sites. It covers two different sets of problems related to such process networks. On the one hand, there is the problem of optimal design of integrated sites, understood as a tightly interconnected network of chemical plants, where the main objective is to guarantee reliable performance at minimum investment cost. This type of problem is dealt with in Chapters 2 and 3. On the other hand, Chapters 4 to 6 deal with planning, scheduling, and simultaneous planning and scheduling problems of multi-site production and distribution networks. After the introductory chapter, Chapter 2 presents a novel MILP formulation for optimal design of integrated sites. The main contribution of this Chapter is the development of an optimization formulation that integrates a model for design and operation of intermediate storage with stochastic flexibility and superstructure optimization in a problem with endogenous uncertainties. Chapter 3 formalizes this model as a two-stage stochastic programming problem with endogenous uncertainties and introduces a novel decomposition algorithm that reduces the number of scenarios and non-anticipativity constraints in the formulation. Chapter 4 presents a new theoretical result to rigorously compare the tightness of temporal vs. spatial Lagrangean decompositions in a class of planning problems; this chapter also presents a useful interpretation of the Lagrange multipliers as transfer prices in the decomposed process network. Chapter 5 introduces a hybrid algorithm for simultaneous planning and scheduling of continuous multi-site production and distribution networks. The algorithm combines bi-level and spatial Lagrangean decompositions. Chapter 6 deals with a continuous-time scheduling formulation that incorporates storage allocation decisions in a setting where there is a limited number of dedicated storage tanks.