Physical Property Modeling of Solvent-Based Carbon Capture Processes with Uncertainty Quantification and Validation with Pilot Plant Data

摘要

The US DOE's Carbon Capture Simulation Initiative (CCSI) has a strong focus on the development of state of the art process models to accelerate the development and commercialization of post-combustion carbon capture system technologies. One of CCSI's goals is the development of a rigorous process model that may serve as a definitive reference for benchmarking the performance of solvent-based CO2 capture systems, using aqueous monoethanolamine (MEA) as a baseline. Among the requirements of this process model is the development of its component submodels (e.g. physical properties) from relevant bench-scale data. Moreover, the process model must take into account parametric uncertainty in the submodels and be validated with both steady-state and dynamic process data collected from a pilot plant over a wide range of operating conditions. This dissertation is focused on two major aspects of the development of this MEA solvent model, namely the development of the physical property models for the MEA-H2O-CO2 system and the validation of the steady-state model with large-scale pilot plant data from the National Carbon Capture Center (NCCC) in Alabama.;The physical property modeling work may be divided into standalone property models and the integrated thermodynamic framework of the system. Viscosity, density, and surface tension models have been developed individually by calibrating parameters, for an empirical model of a given form, to fit experimental data from the open literature. The thermodynamic framework has been developed within Aspen PlusRTM, using the e-NRTL model as a starting point, by regressing model parameters to fit vapor-liquid equilibrium (VLE), heat capacity, and heat of absorption data. A parameter selection methodology using an information criterion has been implemented for reducing the model complexity. A methodology for uncertainty quantification (UQ) has also been included for all property models, in which Bayesian inference is used to update distributions of model parameters in light of experimental data.;The physical property models, along with separately developed mass transfer, hydraulic, and reaction kinetics models, are incorporated into the overall process model. This model has been validated with steady-state data from NCCC for a total of 23 test runs, and the model predictions of absorber and stripper column performance have been shown to match the experimental data with satisfactory fit. The parametric uncertainty from the process submodels are propagated through the process model in order to study the resulting uncertainty in the process variables of the system, notably the CO2 capture efficiency of the absorber and the amount of CO2 regenerated in the stripper. Concurrent sensitivity studies have been performed, which provide insight into the relative contributions of the uncertainty in particular submodels to the overall process uncertainty.;Finally, some ongoing work related to the solvent model project is also presented. In one project, a methodology for scale-up uncertainty quantification is being developed, in which the effect of radial liquid distribution on column performance is estimated and preliminary comparison of this model to process data is made. The final project involves using the completed process model for planning a second MEA campaign at NCCC, which is ongoing at the time of the writing of this dissertation. In this work, the estimated uncertainty in absorber efficiency is quantified as a function of key manipulated variables by propagating the submodel parametric uncertainty through the absorber model over the range of input variables. An initial set of test conditions has been designed with the objective of choosing points for which the estimated uncertainty is relatively high, while maintaining a spread of the conditions throughout the input space. A methodology has been proposed for using Bayesian inference to update the parametric uncertainty as the data are collected.