Carbon Capture and Storage Optimization with Machine Learning

Abstract

This study examines the potential for enhancing carbon capture and storage (CCS) processes by machine learning to markedly improve performance across diverse capture methods, including as absorption, adsorption, membrane separation, and cryogenic distillation. Through the systematic adjustment of critical operating parameters, including temperature, pressure, flow rates, and sorbent characteristics using machine learning algorithms, we saw significant improvements in CO₂ collection efficiency. The use of optimum operating parameters, namely a temperature range of 40-60°C for absorption and a pressure range of 3-5 bar for adsorption, resulted in a 30% enhancement in capture efficiency. Moreover, machine learning models, namely Random Forest and Support Vector Machines (SVM), achieved a maximum enhancement of 20% in forecasting ideal operating parameters for membrane separation and cryogenic systems. Reduced cycle durations in adsorption processes, facilitated by predictive modeling, resulted in a 15% improvement in CO₂ removal rates. The models’ capacity to forecast sorbent regeneration conditions led to a 10% decrease in energy use. Machine learning algorithms adeptly optimized process-specific parameters, including material composition and flow dynamics, enhancing membrane performance by 18% and cryogenic systems by 12%. These results highlight the significance of using machine learning to customize CCS methods for particular materials and situations, facilitating more sustainable, efficient, and scalable carbon capture systems.