CO₂ absorption process in aqueous tri-solvent blended amine systems with experiments and artificial intelligence...

Area of Science:

• Chemical Engineering
• Environmental Science
• Computational Chemistry

Background:

• Existing CO2 capture studies lack clarity on component synergy and multi-factor regulation in ternary mixed amine systems.
• There's a need for robust prediction models applicable to diverse scenarios in CO2 absorption.
• Ternary mixed amine systems offer potential for efficient carbon capture but require detailed mechanistic understanding.

Purpose of the Study:

• To elucidate the synergy mechanism and quantify multi-factor regulation laws in MEA/MDEA/PZ and MEA/AMP/PZ systems for CO2 capture.
• To establish high-efficiency machine learning models for predicting CO2 absorption performance.
• To optimize component ratios and process parameters for enhanced CO2 absorption.

Main Methods:

• Experimental investigation of CO2 absorption performance under varying molar ratios, PZ concentrations, temperatures, and gas flow rates.
• Development and comparison of eight machine learning models (supervised, ensemble, deep learning) using 20,000 experimental data points.
• Feature selection (absorption time, temperature, amine concentrations) and model optimization using data standardization and stratified random splitting.

Main Results:

• 1.25 mol/L PZ demonstrated synergistic effects, enhancing CO2 absorption rate and capacity.
• Optimal parameters identified: 2.25 mol/L MEA + 1.5 mol/L MDEA + 1.25 mol/L PZ achieved 5.48 × 10^-5 mol/s absorption rate and 0.496 mol capacity.
• XGBoost and LSTM models exhibited superior predictive performance with average errors below 0.05, significantly outperforming SVR.

Conclusions:

• Ternary mixed amine systems, particularly with optimized PZ concentrations, show significant potential for CO2 capture.
• Machine learning, specifically XGBoost and LSTM, provides accurate and efficient tools for predicting CO2 absorption performance.
• Quantified factor influence weights advance the understanding and design of advanced CO2 capture technologies.