(535a) Modeling and Optimization of a MOF-Based Pressure/Vacuum Swing Adsorption Process Coupled With Material Selection

The huge amount of carbon dioxide emitted every year has raised significant concern about global climate change. Extensive attention has been paid to Carbon Capture and Sequestration (CCS) to reduce CO₂ emission from large, centralized sources such as power plants. Pressure or vacuum swing adsorption (PSA/VSA) is a promising technology for carbon capture due to its relatively better separation performance, lower energy consumption and lower cost [1]. In order to achieve good performance with PSA/VSA, process optimization and material selection are two important considerations. Process variables, like adsorption and desorption pressures, and material properties, like the isotherm shape and heat of adsorption, can significantly influence process performance. Most previous works have considered these two aspects separately. They either optimize process variables with a fixed material [2, 3], or select materials independent of detailed process modeling [4, 5]. This separated approach has the drawbacks that the performance of a process configuration can be limited by improper choice of adsorbent and that material evaluation without a process model can lead to a non-optimal selection. In this work, we integrate material selection into process modeling by adding integer variables to model material selection. This will enable us to simultaneously choose materials as well as optimize process performance by solving a mixed integer dynamic optimization (MIDO) problem. Our objective is to maximize CO₂ purity and recovery at the same time.

A standard Skarstrom PSA/VSA cycle is modeled with the introduction of integer variables to model the selection of adsorbent materials from a number of promising candidates, such as zeolite 13X, zeolite 5A, ZIF-78, Ni-MOF-74, etc. In a Skarstrom cycle, there is a conflict between purity and recovery. In order to achieve good separation performance, we have two objectives:  maximizing CO₂ recovery and purity at the same time. The resulting problem is a partial differential equations (PDEs) constrained multi-objective MIDO program. Multi-objective optimization can be solved using the ɛ-constraint method, which generates a Pareto-optimal curve by sequentially solving a set of subproblems with different ɛ values. The governing partial differential equations are discretized completely both in time and space. We apply a finite volume method to discretize the spatial domain and a Radau collocation scheme for temporal discretization. This transforms the MIDO problems to mixed-integer nonlinear programs (MINLP)which can be solved using general-purpose MINLP solvers.

Optimization is performed for several candidate materials used for PSA/VSA purposes.  The accuracy of the optimal solutions are verified using MATLAB simulation in which the PDEs are only discretized in space and resulting ODEs are solved using ODE15s in MATLAB. Optimal solutions show the optimal operating conditions and best separation performance for the PSA/VSA cycle. Meanwhile, we identify the best material for the process. The Pareto-curve helps us understand the tradeoff between decision variables under different recovery requirements.