Campaign optimization of multicomponent reactive batch distillation

When resource utilization and/or minimization of waste are comparable in importance to the production rate, the design of the operation policy should encompass an entire campaign of batches rather than a single batch. This notion of campaign optimization is particularly relevant to reactive batch distillation which produces significant amounts of off-cuts; reprocessing these off-cuts merely based on consideration of a single batch may lead not only to inefficient production rates but also to the inefficient utilization of reactants. Reactive Distillation can be classified into two categories. The first involves equilibrium limited reversible reaction. For these applications, the concept of distillation characteristic is introduced and its exploitation is shown to result in a simple but effective reprocessing policy for off-cuts. As an extension of this analysis, a two-level optimization formulation is proposed which optimizes net profit from the entire operation of reactive distillation. The second category involves irreversible reaction and a side product. A two-stage process has been suggested for these types of applications. The economic benefits of two-level optimization and two-stage process are demonstrated with case-studies. For the purpose of this study, RBDOPT (Reactive Batch Distillation Optimization module) has been developed that provides a general framework for simulation and optimization of multicomponent reactive batch distillation. It models the distillation process as a DAE (differential algebraic equation) system and can handle various modes of operation such as the continuous flow of one of the reactants, intermediate feed, inverted distillation etc. Physical estimates are obtained from PPDS that supports a large number of species and routes to model non-ideal behavior. Coded in an object-oriented environment (C++) RBDOPT uses distributed computing techniques to achieve a speedup almost equal to the number of control variables while solving the optimal control problem