Nonlinear Model Predictive Control of Multicomponent Distillation Columns Using Wave Models

Abstract

A novel control concept for multicomponent distillation columns is presented. The concept is based on nonlinear wave-propagation phenomena that occur in counter-current separation processes. On this basis a reduced order model has been developed in previous work that not only considers profile positions but also the profile shape itself. The reduced model gives direct access to key parameters of the plant, such as the separation front positions. Furthermore, it allows realtime computations for multicomponent distillation columns. Such a model is used for both, the nonlinear model predictive control (NMPC) and the observer design. The observer uses temperature measurements and gives estimated temperature and concentration profile positions as well as compositions in the product streams. The robustness of the observer is shown intuitively and in simulation studies. The control of multicomponent distillation is formulated within the NMPC framework by penalising the deviation of the front positions from their reference points and ensuring the product specifications by means of constraints. By directly taking account of product specifications the presented control concept differs from inferential control schemes known from literature. Due to the fact that the concept is based on simple temperature measurements an industrial application seems easily possible.