Aspen modeling for MEA-CO₂ system: Solution approach for rate-based column model

This paper deals with modeling of post-combustion CO₂ removal by monoethanolamine (MEA) absorption-stripping loop with the use of a rate-based model. In comparison with the equilibrium-based stage model, the rate-based model will not only provide overall column performance, such as the capture efficiency and the rich solvent carbon loading (C/N), but also provide in detail column profiles, such as temperature, concentration, absorption rate, and etc. For process simulation, obtaining the accurate profiles is essential when the rate-based model is applied. In terms of rate-based model, due to reaction or absorption coupled with temperature and concentration, the rate varies along the column and through the liquid film, which can potentially lead to high gradients in concentrations and temperature, such as at near the interface of liquid film and near the top of the absorber column. When solving absorption rate with high gradients, obtained results or profiles may be affected by the setup of numerical resolution, i.e. solution approach. To address the solution approach, the impact of stage size and film discretization have been evaluated. The resulting simulated profiles shift in correspondence to the resolution of stage sizing and film discretizing. Divergence occurs in simulation when the resolution does not match the gradient. For example it has been found that a finer stage size than the theoretical stage HETP is required to reach the correct column profile [2]. The same happens for the film discretization, where very fine cells are required near the interface for the fast reaction within the liquid film, which is directly related to the local and chemical reaction enhanced mass transfer coefficient. Successful modeling of the MEA-CO₂ loop relies not only on how the model is applied, but also on whether and how the gridding for numerical solution is carried out. This paper shows modeling results obtained with different setup of numerical resolutions to demo its impact on the performance. The variation of results demonstrated in this paper may partly explain the profile difference in results between different software and model setup [1].