Modeling and Computational Strategies for Nonlinear discrete Optimization in the Synthesis and Planning of Process Systems

摘要

This presentation provides an overview of recent developments in nonlinear discrete optimization. There include, systematic superstructure representations, generalized disjunctive programming, global optimization, and bilevel decomposition techniques. As will be shown a more fundamental understanding of superstructures has been achieved with the use of the State-Task-Network and State-Equipment-Networks. Generalized disjunctive programming, which correspond to optimization problems involving discrete and continuous variables, and are subject to algebraic equations, disjunctions and logic propositions, provides a more natural and direct way of modeling problems. Global optimization techniques for structured problems that involve bilinearities, linear fractional and concave terms, can rigorously guarantee global optimality, and be solved in many cases with reasonable computation time. Finally, bilevel decomposition techniques provide an effective approach for tackling large-scale planning problems by developing higher level aggregate models that determine the major types of discrete decisions. We discuss the basic ideas in each of the four developments cited above, and present example problems that include the synthesis and planning of separation and energy systems, process flowsheets, batch plants, infrastructure planning of oilfields, and scheduling for new product development.