(744c) Nonlinear Model Predictive Control of Air Separation Processes Based on Model Reduction

Nonlinear model predictive control (NMPC) of chemical processes is based on real-time solution of dynamic optimization problems. To enable NMPC for large-scale processes, such as cryogenic air separation units (ASU), model reduction techniques can be applied [8]. These can drastically reduce the computational time while yielding an accurate solution of the dynamic optimization problem. Recently, we have proposed two approaches to reducing dynamic models of multi-component distillation columns [2, 6]. The transient wave propagation approach [2] builds on the geometrical wave phenomenon observed in distillation columns and extends the applicability of existing theory [4] to load change scenarios. Data-driven hybrid compartment modeling [6] combines the compartment concept [5] and substitution of mathematically complex algebraic model blocks by artificial neural networks. The two strategies have been applied for economic NMPC of an ASU benchmark process including a single-section column.

Industrial air separation processes, however, feature complex flowsheets including multi-sectional columns with several feed and withdrawal streams [3]. Reduced dynamic modeling and NMPC for these ASU topologies has not yet been investigated in the literature. In this work, we implement an NMPC for an ASU of industrial complexity and demonstrate its performance in an in-silico closed-loop control case study. To this end, we perform model reduction to derive a controller model suitable for real-time applications. We implement the NMPC with Extended Kalman Filter in Python and use our open-source optimization framework DyOS [1] for solving the nonlinear optimal control problems. We compare an NMPC implementing a reduced model with a controller employing mechanistic stage-by-stage modeling. The obtained CPU times indicate that the proposed framework provides a real-time capable control strategy for an industrial ASU.

References

[1] A. Caspari, A. Bremen, J. M. Faust, F. Jung, C. D. Kappatou, S. Sass, Y. Vaupel, R. Hannemann-Tamás, A. Mhamdi, and A. Mitsos. DyOS – A framework for optimization of large-scale differential algebraic equation systems. In Computer Aided Chemical Engineering, volume 46, pages 619–624. Elsevier, 2019.
[2] A. Caspari, C. Offermanns, A. M. Ecker, M. Pottmann, G. Zapp, A. Mhamdi, and A. Mitsos. A Wave Propagation Approach for Reduced Dynamic Modeling of Distillation Columns: Optimization and Control. Under Review, 2020.
[3] H.-W. Häring and A. Belloni. Industrial gases processing. John Wiley & Sons, 2008.
[4] A. Kienle. Low-order dynamic models for ideal multicomponent distillation processes using nonlinear wave propagation theory. Chemical Engineering Science, 55(10):1817–1828, 2000.
[5] J. Lévine and P. Rouchon. Quality control of binary distillation columns via nonlinear aggregated models. Automatica, 27(3):463–480, 1991.
[6] P. Schäfer, A. Caspari, K. Kleinhans, A. Mhamdi, and A. Mitsos. Reduced dynamic modeling approach for rectification columns based on compartmentalization and artificial neural networks. AIChE Journal, 65(5):e16568, 2019.
[7] P. Schäfer, A. Caspari, A. Mhamdi, and A. Mitsos. Economic nonlinear model predictive control using hybrid mechanistic data-driven models for optimal operation in real-time electricity markets: In- silico application to air separation processes. Journal of Process Control, 84:171–181, 2019.
[8] D. R. Vinson. Air separation control technology. Computers & Chemical Engineering, 30(1012):1436–1446, 2006.

Acknowledgements

The authors gratefully acknowledge the financial support of the Kopernikus project SynErgie by the Federal Ministry of Education and Research (BMBF) and the project supervision by the project management organization Projektträger Jülich.