Machine learning for industrial processes: Forecasting amine emissions from a carbon capture plant

Area of Science:

• Environmental Science
• Chemical Engineering
• Machine Learning Applications

Background:

• Amine-based carbon capture processes are crucial for reducing CO2 emissions.
• A significant environmental impact is the emission of amine solvents into the atmosphere.
• Intermittent power plant operation introduces operational complexities affecting these emissions.

Purpose of the Study:

• To investigate the effect of intermittent operation on amine solvent emissions.
• To develop a predictive model for amine emissions using machine learning.
• To evaluate the impact of various interventions on emission mitigation.

Main Methods:

• Conducted stress tests on a carbon capture plant using a mixed amine solvent (2-amino-2-methyl-1-propanol and piperazine, CESAR1).
• Developed and applied a machine learning model to forecast emissions.
• Analyzed the effects of different interventions on individual solvent component emissions.

Main Results:

• Machine learning model successfully forecasted emissions under intermittent operation.
• Identified interventions with opposing effects on the emissions of different amine components.
• Demonstrated that standard mitigation strategies may be inadequate for mixed-amine systems.

Conclusions:

• Mitigation strategies for single-amine systems require re-evaluation for mixed-amine carbon capture plants.
• Machine learning offers a powerful approach for modeling complex emission processes.
• The developed approach has potential for broader application in process modeling and emission control.