Absorption of Carbon Dioxide - Modeling and Experimental Characterization
Abstract
Recent studies on the costs of different processes for CO2 capture from power plants show that post-combustion CO2 removal by absorption with chemical solvents has the potential to remain the most important process also in the future. However, technological improvements are necessary to lower the high capital cost and energy requirements of the absorption process. It is expected that process design innovations or the use of better solvents can reduce the capital and energy costs. Extensive research activities are directed towards the optimization of the process and also the search for new environmentally friendly and energy efficient absorbents. Reliable models are needed for testing of absorber performance and/or the effect of various absorbent properties.
In the first part of this thesis, the finite volume method (FVM), typically used in CFD-codes, was adopted for the first time to develop and implement a model for a membrane gas absorber module. This was done to provide a basis for further implementation of highly reactive systems in CFD, a problem that has not been focused in this area. The FVM based model is flux conservative and can more easily be expanded to other geometries than tubes. The developed model was compared to a validated finite difference based model. The results were found to be very similar, but the finite volume method gave slightly lower absorption fluxes. The model provided reasonable system variables profiles and predictions of absorption rates as well as the effects of varying operation conditions. However, the model performance was not optimal and further improvements are required regarding the efficiency of the algebraic solver, numerical grid optimization, and submodels for physical properties, reaction kinetics and equilibria. This part of the work was published as part of Hoff et. al.(2003).
In the second part of the thesis vapor-liquid equilibria (VLE) were studied. A VLE model was implemented for the binary and ternary CO2/H2O/MEA/ MDEA systems. For the activity coefficients, the ElecGC model by Lee (1996) that combines the group contribution method UNIFAC with the mean spherical approximation method (MSA) was adopted and modified by exchanging the MSA with the simpler extended Debye-Hückel model (DHext). The proposed UNIFAC + DHext model differs from the electrolyte UNIFAC models (UNIFAC+ DH) presented in the literature. It represents a new approach in the way the group contribution method and the electrolyte Debye-Hückel theory were combined. The parameters of the model were fitted to experimental binary and ternary VLE data. The fits obtained with the original model by Lee (1996) and the UNIFAC + DHext model were of the same overall quality and the calculation time for the UNIFAC + DHext model was much shorter. The average relative deviations of 22% for MEA and 27% for MDEA were comparable to the fit of other models reported in the literature. The deviations are believed to be caused to a large extent by the inconsistencies and scatter in the experimental data. It was concluded that for a reliable parameter estimation a database needs to be established based on thorough evaluation of the available experimental data sets. It was also concluded that VLE data alone are not sufficient to accurately determine the model parameters and to judge the quality of the model. For the evaluation as well as for further refinement of the VLE models, additional information about the liquid phase compositions are needed.
The liquid phase composition was investigated in the third part of the thesis. 13C NMR studies were performed investigating the liquid phase composition in samples where various amounts of CO2 were dissolved in aqueous alkanolamines (BEA, MDEA, and MEA) at varying temperatures. Chemical shifts of functional groups present in the systems were reported. The liquid phase compositions were calculated for 20 and 40oC with an estimated accuracy of the species concentrations of about 4-5%. The obtained speciations were based on the NMR spectra only and represent thereby additional and independent information on the systems. The experimental speciation was compared with the speciation predicted by the model. The comparisons showed qualitative agreement, but quantitatively the agreement was not satisfactory. The model parameters proved to be very sensitive to the speciation data and refitting of the parameters based on molecule and ion compositions were performed. However, it was found that more experimental data for the free CO2 concentrations are needed.
At higher temperatures, the spectra were broadened by fast chemical exchanges between the species. The dynamic nature of the system complicates the quantitative evaluation of the spectra, but can provide more information about the energy of chemical bonds and thus the reaction rates. Dynamic simulations of the spectra at higher temperatures are planned to be performed.