(170a) Nonlinear Distributed Model Predictive Control of Gas Sweetening Processes

Gas sweetening is one of the key processes in the production train of liquefied natural gas and in the post-combustion flue gas treatment, where acid gases like CO₂ and H₂S are removed through selective absorption in solvents (e.g., diethanolamine, potassium carbonate, etc.) to reduce their release into the environment [1, 2]. A typical gas sweetening plant comprises single or multi-stage absorber columns for the absorption of acid gases into the solvent, and stripper units to regenerate the spent solvent for recycle. Material recycles and heat integration in the plant are essential for minimization of energy and utility waste. Many efforts have been made to design, model and optimize the gas sweetening plants, but fewer research efforts have been dedicated to controlling the process. Single stage absorption-regeneration has been widely studied from the control viewpoint, e.g. control under feed flow rate fluctuations [3], decentralized control of the absorber and stripper column under disturbances [4, 5], self-optimizing control for determining the control configuration [6], and single-input-single-output PID control [7, 8]. In these studies, the interaction effects due to integration between process units have not been considered [9]. These become important in the context of transient operation of such plants. In addition, the gas sweetening plants are frequently subject to disturbances in the form of fluctuations in the mass flow rate of the acid gas or the acid content of the feed. These disturbances can propagate through the integrated plant, making its control challenging. Motivated by these, this work addresses the nonlinear optimization-based control of such plants.

The optimization-based control of gas sweetening process units and their interconnections, which can be described by a set of algebraic and differential equations (ODEs and PDEs), requires solving a large-scale constrained nonlinear dynamic optimization in real-time [10], which is computationally expensive. In this work, we implement a distributed model predictive control (DMPC) architecture comprising local controllers with some level of cooperation and communication [11, 12]. A prerequisite for the application of such a strategy is the decomposition of the large-scale system into the corresponding subsystems. To this end, we adopt a recently developed framework that determines optimal decompositions based on community detection methods [13-15]. The resulting decompositions maximize the modularity of the corresponding subsystems, thereby minimizing the interactions among them.

Specifically we implement an iterative DMPC architecture to address the output regulation problem. The sweetening process consists of two absorbers, two regenerator columns, and two heat exchangers which are tightly integrated. The optimal decomposition separates the first-stage absorption and regeneration columns from those in the second stage, recommending a distributed control structure of two local controllers which can communicate over the network. The local controllers are developed based on the corresponding ODE systems and coarse discretizations of the PDEs describing the absorber/stripper columns. The closed-loop performance and the average computation time are evaluated using the detailed process model implemented in gPROMS. We additionally compare the results with those of centralized and fully decentralized MPC synthesis.

[1] A. Bedelbayev, T. Greer, B. Lie, Model based control of absorption tower for CO_2 capturing, Proceedings, 49th International Conference of Scandinavian Simulation Society (SIMS 2008), October 7th-8th 2008, Oslo University College, Oslo, Norway. ISBN-13: 978-82-579-4632-6, 2008.

[2] M. Abu-Zahra, J. Niederer, P. Feron, G. Versteeg, CO_2 capture from power plants Part II: A parametric study of the economical performance based on mono-ethanolamine, International Journal of Greenhouse Gas Control, 1, 135–142, 2007.

[3] A. Lawal, M. Wang, P. Stephenson, G. Koumpouras, H. Yeung, Dynamic modelling and analysis of post-combustion CO₂ chemical absorption process for coal-fired power plants, Fuel,  89(10), 2791–2801, 2010.

[4] S. Ziaii, G. Rochelle, T. Edgar, Dynamic modeling to minimize energy use for CO₂ capture in power plants by aqueous monoethanolamine, Industrial & Engineering Chemistry Research, 48 (13), 6105–6111, 2009.

[5] H. Kvamsdal, J. Jakobsen, K. Hoff, Dynamic modeling and simulation of a CO_2 absorber column for post-combustion CO2 capture, Chemical Engineering Process, 48 (1), 135–144, 2009.

[6] M. Schach, R. Schneider, H. Schramm and J. Repke, Control structure design for CO_2 absorption processes with large operating ranges, Energy Technology, 1, 233 – 244, 2013.

[7] C. Biliyoka, A. Lawal, M. Wanga, F. Seibert, Dynamic modelling, validation and analysis of post-combustion chemical absorption CO_2 capture plant, International Journal of Greenhouse Gas Control, 9, 428–445, 2012.

[8] Y. Lin, T. Pan, D. Wong, S. Jang, Y. Chi, C. Yeh, Plantwide control of CO_2 capture by absorption and stripping using monoethanolamine solution, Industrial & Engineering Chemistry Research, 50, 1338 – 1345, 2011.

[9] M. Wang, A. Lawal, P. Stephenson, J. Sidders, C. Ramshawa, Post-combustion CO_2 capture with chemical absorption: A state-of-the-art review, Chemical Engineering Research and Design, 89, 1609–1624, 2011.

[10] R. Lopez-Negrete, F.J. Damato, L.T. Biegler, and A. Kumar. Fast nonlinear model predictive control: Formulation and industrial process applications, Computers & Chemical Engineering, 51, 55–64, 2013.

[11] D.Q. Mayne, Model predictive control: Recent developments and future promise, Automatica, 50(12), 2967–2986, 2014.

[12] P.D. Christofides, R. Scattolini, D. Munoz de la Pena, and J. Liu. Distributed model predictive control: A tutorial review and future research directions, Computers & Chemical Engineering, 51, 21–41, 2013.

[13] S. Heo and P. Daoutidis. Control-relevant decomposition of process networks via optimization-based hierarchical clustering, AIChE Journal, 62(9), 3177–3188, 2016.

[14] S. Heo, W.A. Marvin, and P. Daoutidis. Automated synthesis of control configurations for process networks based on structural coupling, Chemical Engineering Science, 136, 76–87, 2015.

[15] M. Moharir, L. Kang, A. Almansoori, and P. Daoutidis. Control relevant network decomposition of distributed parameter system networks, Computers & Chemical Engineering, 2017, submitted.