Experiment and Modeling Using Machine Learning Application of CO2 Capture in a Rotating Packed Bed Pilot Plant of One TPD Capacity

Abstract

Carbon dioxide (CO2) is the primary factor responsible for the rise in the average temperature of Earth's atmosphere. The primary source of its release into the atmosphere is the combustion of fossil fuels and the emissions from industrial activities. To stabilize atmospheric CO2, Carbon Capture and Utilization will play a significant role shortly, and chemical absorption/solvent-based technology can be considered as one option for CO2 capture. This work involves the absorption of CO2 in a process intensified Rotating Packed Bed (RPB) using monoethanolamine (MEA) based solvent as a base case. The CO2 absorption process in RPB encompasses operating variables, such as rotating speed, solvent concentration, liquid and gas flow rate, temperature., etc. By modeling this process, one can gain insight into the behavior and pattern of CO2 absorption and the impact of various operating parameters. This study utilizes Artificial Neural Networks (ANN), a subfield of machine learning (ML), and ML algorithms to predict the process of CO2 absorption in RPB. ML models are versatile instruments that can be employed to simulate and forecast highly non-linear phenomena. Experiments have been conducted in a newly built pilot plant RPB absorber. The effect of liquid and gas flow rates, RPM, L/G ratio, CO2 partial pressure, etc., on the %CO2 removal of inlet CO2 in this newly built pilot plant are presented here. The CO2 absorption efficiency has been modeled using a Rectified Linear Unit (ReLU) transfer function, and the findings reveal that the root mean square error (RMSE) value is less than 0.5. Hence, the machine learning method accurately predicts the CO2 absorption in MEA solvent to a satisfactory degree.

Keywords: CO2 capture, chemical absorption, rotating packed bed