CO2 solubility in aqueous 1-methylpiperazine and its blend with 2-amino-2-methyl-1-propanol: New experimental data and leveraging Random Forest model for accurate prediction

A new study of amines with enhanced properties is crucial for its use as an appropriate solvent for CO₂ absorption process. This is essential to overcome current financial and environmental obstacles to implementing large-scale CO₂ capture. In this work, the identified amines to formulate a good aqueous solvent mix are 1-methylpiperazine (1-MPZ) and 2-amino-2-methyl-1-propanol (AMP). 1-MPZ has a higher thermal resistance and lower CO₂ regeneration cost, which makes it a potential candidate for the process. This paper presents the results of experiments on the equilibrium solubility of CO₂ in aqueous 1-MPZ solutions and its blends with AMP along with new vapor pressure data for aqueous 1-MPZ and its blend with AMP. The study examined CO₂ solubility at 0.10, 0.20, 0.30, and 0.40 wt fraction (w) of 1-MPZ and temperatures of 303.15, 313.15, 323.15, and 393.15 K. The CO₂ solubility experiments for the new blended mixture of AMP + 1-MPZ were also performed in this study. Four different sets were prepared by varying the individual concentrations of 1-MPZ and AMP; keeping the total amine concentration 0.40 w for the same temperature range. The experimental results indicate that aqueous 1-MPZ exhibits good cyclic capacity for CO₂ intake which further enhances when combined with AMP. Results indicate that aqueous 1-MPZ exhibits high CO₂ solubility of 0.78 to 1.01 moles of CO₂/moles of amine at normal absorber conditions of temperature and composition. Again, the addition of 1-MPZ to AMP has increased the cyclic capacity. The blended aqueous amine (AMP+ 1-MPZ) shows a cyclic capacity of 118 g CO₂/kg-solution whereas aqueous 1-MPZ of the same amine strength shows only 81 g CO₂/kg-solution. Hence, an increase of about 45% cyclic capacity is achieved with the blended solvent. The combination of 1-MPZ+AMP allows for improved performance in terms of both solubility and regeneration efficiency. The study also incorporated Random Forest (RF) machine learning model to predict CO₂ solubility based on parameters like temperature, concentration, and CO₂ partial pressure. The RF model demonstrated excellent prediction accuracy with R² values of 0.98 and 0.99 for both solutions and least RMSE values of 0.03 and 0.02 for 1-MPZ and (1-MPZ +AMP) blend respectively.