(202a) Optimal Design and Control of Reactive Distillation Systems Under Uncertainty
AIChE Annual Meeting
Monday, November 11, 2019 - 3:32pm to 3:55pm
The concept of distillation column superstructures has been successfully applied in the past for the optimal design of simple distillation columns (Dünnebier, G. and Pantelides, 1999). This network representation method determines all the desired column design and operation specifications by solving a mathematically complex optimization problem. The objective function is typically a cost function, and there are constraints imposed either by the system itself (mass balances) and/or the user (product specifications). Panjwani and co-workers (Panjwani et al., 2005) applied the superstructure concept for the optimal design (structural decisions) of a single reactive distillation column. In this work, the concept of distillation column superstructures will be extended to general reactive distillation systems in order to determine both structural and operational decisions; including but not limited to, total number of stages, feed stages location, existence of side draw streams, reactive zones etc. In addition to a single reactive distillation column, the synthesis of more complex reactive distillation processes, e.g. including a pre-reactor, a sequence of reactive/non-reactive distillation columns etc., will also be considered.
The optimization problem considered is a Mixed-Integer Non-Linear Problem (MINLP), as the design variables are both continuous (e.g. reflux ratio) and discrete (e.g. number of stages). The solution of the MINLPs firstly determines the existence of the units and secondly, their design and operating parameters in case they exist. In this way, the optimal design of various process alternatives can be determined, thus allowing the selection of the most suitable process alternative based on the given objective function. The software used in this work is gPROMS ProcessBuilder (Process Systems Enterprise, 2019), selected due to its rigorous mathematical solvers and excellent optimization capabilities.
Different types of reactive distillation processes, with varying difficulty in terms of the separation task (easy/difficult separation) and different kinetics (fast/slow reaction), are considered. The MINLP considered for each case aims to minimize the total annualised (capital and operational) cost of the process, subject to constraints such as required product purity. Using the superstructure approach, the optimal design parameters for each case are determined and the impact of the process characteristics on the final design is investigated.
Once the optimal steady-state configuration has been determined, the optimal design is then processed in dynamic mode in order to investigate the controllability and stability of the process by considering various disturbances which could potentially be introduced (e.g. feed flow rate disturbance). This investigation will show how the optimal steady state design will perform during operation, and will establish whether any adjustments are needed to the design in order to render it more stable and/or controllable.
The optimal design is obtained by the rigorous optimization of a mathematical model formulated as an MINLP. This model, however, depends on a number or model parameters, for which the reaction parameters are key for reactive distillation. A robust design should tolerate a certain level of parameter uncertainty, and for this reason, the impact of uncertainty in the reaction kinetic parameters on the optimal design will also be investigated in order to determine exactly how sensitive the optimal designs are to these parameters.
It will be demonstrated how the separation parameters (given by the relative volatilities range) impact on the process synthesis and the reaction. Furthermore, it will be demonstrated that the existence of a slow or fast reaction within the column also has an impact on the structural and operational decisions for the design of a reactive distillation column. Of particular interest is the interaction between ease of separation and the characteristics of the reaction system, and it will be shown that the process with a slow reaction and more difficult separation has the highest total cost due to its more demanding process design & operation. Moreover, the dynamic analysis will demonstrate how reaction kinetics and separation parameters impact on the stability and controllability of a process, and how controllability can determine whether a process alternative is actually viable or not. The uncertainty analysis will also demonstrate how sensitive the optimal design is to input uncertainty, and will provide insight into how uncertainty may render an optimal design practically infeasible.
In conclusion, this work will provide a methodology for how to select the most suitable reactive distillation process design that is robust under dynamic operation as well as under input uncertainty, based on various (simple and more complex) process configuration alternatives as expressed by the superstructure.
Dünnebier, G. and Pantelides C.C. (1999). Optimal Design of Thermally Coupled Distillation Columns, Ind. Eng. Chem. Res., 38, 162-176.
Harmsen, G. J. (2007). Reactive distillation: The front-runner of industrial process intensification: A full review of commercial applications, research, scale-up, design and operation, Chem. Eng. Process., 46, 774â780.
Kiss, A. A. (2013). Advanced Distillation Technologies: Design, Control and Applications, Wiley, Chichester.
Luyben, W. L. and Yu, C. C. (2008). Reactive Distillation Design and Control, Wiley, New Jersey.
Panjwani, P. et al. (2005). Optimal design and control of a reactive distillation system, Eng. Optimiz., 37, 733-753.
Process Systems Enterprise (2019), gPROMS, www.psenterprise.com/gproms.
Wang, S. J. et al. (2010). Plantwide Design of Ideal Reactive Distillation Processes with Thermal Coupling, Ind. Eng. Chem. Res., 49, 3262-3274.