Estimating Carbon Dioxide Solubility in Brine Using Mixed Effects Random Forest Based on Genetic Algorithm: Implications for Carbon Dioxide Sequestration in Saline Aquifers

Accurate prediction of carbon dioxide (CO2) solubility in brine is crucial for the success of carbon capture and storage (CCS) by means of geological formations like aquifers. This study investigates the effectiveness of a novel genetic algorithm-mixed effects random forest (GA-MERF) model for estimating CO2 solubility in brine. The model's performance is compared with established methods like the group method of data handling (GMDH), backpropagation neural networks (BPNN), and traditional thermodynamic models. The GA-MERF model utilizes experimental data collected from literature, encompassing key factors influencing CO2 solubility: temperature (T), pressure (P), and salinity. These data are used to train and validate the model's ability to predict CO2 solubility values. The results demonstrate the superiority of GA-MERF compared to the other models. Notably, GA-MERF achieves a high coefficient of determination (R) of 0.9994 in unseen data, indicating a strong correlation between estimated and actual CO2 solubility values. Furthermore, the model exhibits exceptionally low error metrics, with a root mean squared error (RMSE) of 2x10(-8) and a mean absolute error (MAE) of 1.8x10(-11), signifying outstanding accuracy in estimating CO2 solubility in brine. Beyond its high accuracy, GA-MERF offers an additional benefit-reduced computational time compared to the other models investigated, with 65 seconds. This efficiency makes GA-MERF a particularly attractive tool for real-world applications where rapid and reliable CO2 solubility predictions are critical. In conclusion, this study presents GA-MERF as a powerful and efficient model for predicting CO2 solubility in brine. Its superior performance compared to existing methods and previous literature highlights its potential as a valuable tool for researchers and engineers working on CCS projects utilizing aquifer storage. The high accuracy, low error rates, and reduced computational time make GA-MERF a promising candidate for advancing the development of effective and efficient CCS technologies.