Dynamic optimization of multicomponent distillation processes

Increasing quality requirements and decreasing production periods demand the real-time optimization of simple units as well as of wholeplants in the process industry. In this work a multicomponent distillation process as the most widespread separation technique in process industries is considered. The system consists of two coupled distillation columns with additional pumps, valves, controllers, condensers and one reboiler. The side column needs no reboiler, since it is fed by a vaporside-withdrawal of the main column. Hence, this configuration requires only little energy but is comparably difficult to operate. After a sudden change of the composition in the liquid feed to the main column, optimal trajectories for the input variables have to be computed in limited time in order to guarantee the desired product specifications. The optimization of a rigorous model of the distillation system with a standard SQP algorithm results in an unacceptably high computation time [3]. In principle, there are two complementary possibilities to reduce the computational effort. First, efficient special tailored algorithms like a multiple shooting approach could be used [2]. Second, the model size and complexity could be reduced. Emphasis in this contribution is on the second option. It turns out that a reduction of the rigorous model by means of simplifying assumptions is not sufficient. A further reduction of the system order beyond those model simplifications is needed. For this purpose, a new model reduction approach for multicomponent systems is used based on nonlinear wave propagation theory [1]. As the order of the system can be dramatically reduced in comparison with the simplest rigorous model, the real-time optimization of the input variables becomes possible. In this contribution, the real-time requirements for the process described above are considered. Afterwards, an overview is given over possible model formulations for distillation processes including different rigorous models and the wave model. The potential of these models with respect to dynamic optimization computations is discussed. Finally, computational results for the different models are given and compared. [1] A. Kienle: Reduced models for multicomponent separation processes using nonlinear wave propagation theory. In CHISA’98, paper 109, Praha, Czech Republic, August 23 - 28, 1998. [2] D.B. Leineweber and A.C. Ströder: Parameter estimation and optimal control for dynamic chemical processes. In Keil, Mackens, Voss and Werther (Eds.): Scientific Computing in Chemical Engineering, Springer-Verlag, Berlin, 1996.[3] C.P. Majer: Erweiterung einer Simulationsumgebung für verfahrentechnische Prozesse zur Parameterschätzung, Versuchsplannung und Trajektorienoptimierung. VDI Fortschritt-Berichte Nr. 3/538, VDI-Verlag, Düsseldorf, 1998