Using Multivariate Data Analysis and ATR-FTIR Spectroscopy for Modeling Components Present During CO2 Capture with Amines

This master thesis is a collaboration between Technology Centre Mongstad (TCM) and University of Bergen. The project is to develop a method to accurately predict total inorganic carbon, total alkalinity and density using spectroscopy and multivariate data analysis. These variables can be used to determine the CO2-loading and MEA concentration. The CO2 concentrations in the atmosphere have been increasing since the 19th century; the increase has been affected by anthropogenic CO2 emissions. The most significant source of anthropogenic CO2 is the combustion of fossil fuels, especially in large power plants. The use of post-combustion CO2 capture at large power plants can decrease the amount of emissions drastically. Monoethanolamine (MEA) has been extensively studied as an aqueous solvent to use in CO2 capture and is a good choice for this purpose. Other solvents have not been this extensively studied. Therefore it is not sure which solvent that is the best choice yet. This thesis aims to use multivariate data analysis to create models that can be used for prediction of the compounds in the MEA-solution at different times in the process. Three response variables are chosen, total inorganic carbon (TIC), total alkalinity (TOT_ALK) and density. TIC can be used to find the CO2 concentration, TOT_ALK for finding the amine concentration and the density is correlated to the CO2-loading. The three response variables are predicted using partial least squares (PLS) models, preprocessing of the data is done with extended multiplicative signal correction (EMSC) or Savitzky-Golay differentiation. Outlier detection has been performed with principal component analysis (PCA). The achieved models have good predictive abilities, with small prediction errors and residuals