2025 AIChE Annual Meeting

(264d) Machine Learning Model for Predicting CO? Solubility in Seawater for Mineral Carbonation

Authors

Yongha Park, Korea Institute of Energy Research
Excessive anthropogenic CO₂ emissions are a primary driver of global climate change. In recent years, alongside efforts to reduce emissions from point sources, growing attention has been directed toward negative emissions technologies aimed at directly removing CO₂ from the ambient environment—where the low concentration of CO₂ presents a significant challenge for capture [1]. In addition to the atmosphere, the ocean is increasingly recognized as a global-scale reservoir for atmospheric CO₂ [2]. However, accurately predicting CO₂ solubility in the chemically complex environment of seawater is essential for efficient CO₂ stripping and subsequent utilization [1, 3]. In particular, seawater-based mineral carbonation has emerged as a promising approach for long-term carbon storage. In this process, the dissolution of CO₂ into seawater is a crucial step, as the formation of stable carbonates depends on the availability of dissolved CO₂ species [3].

This study presents a machine learning-based approach for predicting CO₂ solubility in aqueous salt solutions relevant to seawater mineral carbonation. An artificial neural network (ANN) model was developed using a dataset of 2406 experimental measurements, covering a wide range of pressures (0.92–712.31 bar), temperatures (273.15–473.65 K), concentrations of electrostricted water (0–90.12 mol/kgw), and dissolved salt mole fractions (0–25.39 mol%), including supercritical conditions. The ANN model, using only four input variables, achieved high predictive accuracy, with an average absolute relative deviation (AARD) of 4.90%, outperforming a conventional thermodynamic model (AARD 5.44%).

The proposed model demonstrates strong extrapolation capability without overfitting and offers a practical and efficient tool for estimating CO₂ solubility in complex salt solutions. These results highlight the potential of data-driven models in supporting CO₂ capture and storage strategies, particularly in marine environments.

[1] Liu, S. S., Song, J. M., Li, X. G., Yuan, H. M., Duan, L. Q., Li, S. C., ... & Ma, J. (2024). Enhancing CO2 storage and marine carbon sink based on seawater mineral carbonation. Marine Pollution Bulletin, 206, 116685.

[2] Kim, S., Nitzsche, M. P., Rufer, S. B., Lake, J. R., Varanasi, K. K., & Hatton, T. A. (2023). Asymmetric chloride-mediated electrochemical process for CO 2 removal from oceanwater. Energy & Environmental Science, 16(5), 2030-2044.

[3] Jeon, P. R., & Lee, C. H. (2021). Artificial neural network modelling for solubility of carbon dioxide in various aqueous solutions from pure water to brine. Journal of CO2 Utilization, 47, 101500.