(512j) Model Predictive Control Performance Analysis to Driving Force Based Reactive Distillation Columns

Abstract: Integration of design and control for reaction-separation systems has been subject to continuous research over the past many years. For any reactive system,we can define the number of elements as being the difference between the number of components and the number of reactions including inert. Therefore, an elementary reaction of two reactants and one product yields a binary element reactive system. Analogously, reaction between multi components can be expressed with respect to multi elements. For such cases, we choose equivalent binary key elements based on relative volatility. Subsequently, driving force-based design can be drawn from which column design parameters such as, number of column stages, optimal feed location, reflux ratio, reboil ratio can be obtained [1]. Previously, it has been shown that the optimal control is achievable if the column is operated at the point where the driving force is maximum [2]. Such integrated optimal design-control has been demonstrated for proportional-integral (PI) control systems. In this work, we demonstrate the applicability of advanced control systems, i.e. model predictive controller (MPC), to several major driving force based reactive distillation applications. To establish the optimal control at maximum driving force for model predictive control structures, we demonstrate four reactive distillation case studies: 1) Methyl-Tert-Butyl-Ether production from isobutene and methanol, 2) Methyl-Tert-Butyl-Ether production from isobutene and methanol in presence of inert (1-butene), 3) Esterification of methanol and acetic acid, and 4) Toluene disproportionation reaction. It is to be noted that 1 and 4 are single feed binary element reactive systems, 2 is single feed multi element reactive system and 3 is an example of double feed multi-element reactive system. For each reactive distillation system, we have compared the design at the maximum driving force with other design alternatives that are not designed based on the maximum available driving force. At first we have generated the reactive vapor liquid equilibrium (VLE) data using ICAS environment. Using the VLE data design parameters have been determined. After that, we have performed rigorous steady state simulation in ASPEN PLUS environment. The dynamic simulation and corresponding PI control structure performance evaluation have been performed in ASPEN Dynamics. The exported dynamic models to ASPEN Dynamics have been checked for steady state consistency. After the dynamic simulation, the model has been linearized at a reference point, and subsequently, state space model has been obtained. Using the state space representation, model predictive control structure has been implemented to the models. In order to verify the integrated design and control, a multi-objective performance function has been defined which includes the total energy consumption associated with the process, integral absolute error and total input variation.For both PI and MPC control structures, we have controlled the top and bottom compositions of our product(s) of interest, with reflux rate and reboiler duty respectively. It has been readily observed that the design-control solution that is based on the largest driving force, has been able to reject feed disturbances with minimum process upset. Both PI and MPC showed better performance for design-control solution compared to the other design alternatives. Our study also includes a comparative study between PI and MPC. The variation of controlled variables under MPC was less than PI for all cases. The key findings of this study are as follows: 1) design at the maximum driving force has a better controller performance, regardless of the choice of the controller (PI or MPC) compared to any other design which is not at the maximum driving force, 2) Design-control solution has a satisfactory performance not only using controllers at the regulatory levels (PI) but also advanced algorithms such as MPC at the supervisory level, 3) MPC performs better than PI for all cases including design-control solution and design alternatives. This study leads to an important conclusion that RD columns are to be designed based on the maximum driving force in the first place to ensure good control irrespective of controllers whether basic regulatory or advanced supervisory controllers. For the existing plants, an economic analysis will determine whether retrofitting the columns is more beneficial than implementing advanced process control.

References

[1] Lopez-Arenas.T., Mansouri. S. S., Sales-Cruz. M., Gani.R. & Pérez-Cisneros, E. S., A
Gibbs energy-driving force method for the optimal design of non-reactive and reactive
distillation columns, In Computers & Chemical Engineering, 128, p. 53-68, 2019.
[2] Mansouri. S. S., Huusom. J., & Gani.R., Systematic Integrated Process Design and Control of
Binary Element Reactive Distillation Processes”, AIChE Journal., 62, p. 3137-3154, 2016.