Multimodel predictive system for carbon dioxide solubility in saline formation waters.

Area of Science:

• Geochemistry
• Chemical Engineering
• Environmental Science

Background:

• Accurate prediction of carbon dioxide (CO(2)) solubility in brine is essential for the success of carbon sequestration technologies.
• Existing mathematical models vary in accuracy across different temperature, pressure, and salinity conditions (T-P-X).

Purpose of the Study:

• To compare eleven mathematical models for CO(2) solubility in brine.
• To develop a multimodel predictive system (MMoPS) for enhanced prediction accuracy in carbon sequestration applications.

Main Methods:

• Evaluated model goodness-of-fit using 173 published CO(2) solubility measurements across wide T-P-X ranges.
• Assessed model performance using statistical methods like AIC and BIC.
• Developed a machine learning classification tree to guide MMoPS model selection.

Main Results:

• Different models showed optimal performance in distinct T-P-X subranges.
• The developed MMoPS effectively selected the most accurate model for given conditions.
• Stratified 5-fold cross-validation confirmed MMoPS's superior performance over individual models.

Conclusions:

• MMoPS significantly increases the overall accuracy of CO(2) solubility prediction for carbon sequestration.
• The developed system provides a robust tool for optimizing carbon capture and storage operations.
• This approach enhances the reliability and efficiency of geological carbon sequestration.