Thermodynamic modeling and uncertainty quantification of CO2-loaded aqueous MEA solutions

Abstract

The accurate characterization of the thermodynamic models of a reactive solvent-based CO2 capture system is essential for adequately capturing system behavior in a process model. Moreover, uncertainty in these models can significantly affect simulation results, although it is often neglected. With this incentive, a model of thermodynamic properties, including vapor-liquid equilibrium (VLE), enthalpy, and solution chemistry, has been developed using a rigorous methodology for the aqueous monoethanolamine (MEA) system as a baseline. The final thermodynamic framework consists of both a deterministic and stochastic model. The deterministic model is developed by regressing parameters of the e-NRTL activity coefficient model in Aspen Plus® to VLE, heat capacity, and heat of absorption data while downselecting the model’s large parameter space through use of information-theoretic criteria. The stochastic model is developed through an uncertainty quantification (UQ) procedure, using the results of the deterministic regression to determine a prior parameter distribution. This prior distribution and the experimental VLE data are used to derive a posterior distribution through Bayesian inference, which is used to represent the final stochastic model. A reaction model in which the kinetics are written in terms of the reaction equilibrium constants, to ensure consistency with the thermodynamics model, is developed. These new thermodynamic and reaction models are incorporated into an existing MEA system process model from the open literature, and the prior and posterior parameter distributions are propagated through the models. This provides valuable insight into the extent to which uncertainty in thermodynamic models affects key process variables, including CO2 capture efficiency and energy requirement for solvent regeneration.