In the relentless battle against global warming, scientists are turning to innovative solutions to mitigate the impacts of climate change. One promising avenue is carbon capture and storage (CCS), where CO2 is injected into saline aquifers deep underground. This method not only reduces atmospheric CO2 levels but also leverages the natural properties of these geological formations to trap the gas securely. However, the dynamics of CO2 dissolution in brine—critical for long-term storage—remain complex and challenging to model accurately.
Enter Amin Shokrollahi, a researcher from the School of Chemical Engineering at the University of Adelaide. Shokrollahi and his team have developed a groundbreaking two-step explainable artificial intelligence (XAI) framework to predict the density of CO2-dissolved brine in geological formations. Their work, published in the journal ‘Inventions’ (translated to English as ‘Inventions’), could revolutionize how we approach CO2 storage in saline aquifers.
The study focuses on solubility trapping, where CO2 dissolves into brine, altering its density and initiating density-driven convection. This process enhances CO2 migration and storage efficiency. “Accurate modelling of these density changes is essential for optimizing CO2 storage strategies and improving long-term sequestration outcomes,” Shokrollahi explains. “Our XAI framework provides a clear and interpretable model that can significantly enhance our understanding and prediction capabilities.”
The research utilized a comprehensive dataset comprising 3,393 samples from 14 different studies, capturing a wide range of brine compositions and salinities. The team employed a two-step modelling approach using random forest (RF) models. The first model predicted the brine volume without dissolved CO2, while the second model predicted the impact of CO2 dissolution on the brine’s volume. Feature importance analysis and SHapley Additive exPlanations (SHAP) values provided interpretability, revealing the dominant role of temperature and ion mass in the absence of CO2 and the significant influence of dissolved CO2 in more complex systems.
The model’s predictive performance was impressive, with R² values of 0.997 and 0.926 for brine-only and CO2-dissolved solutions, respectively. This high accuracy is a testament to the robustness of the XAI framework and its potential to shape future developments in the field.
The implications of this research are vast. For the energy sector, accurate modelling of CO2-dissolved brine density could lead to more efficient and cost-effective CCS strategies. By understanding the dynamics of density-driven convection, engineers can optimize injection rates and storage sites, reducing the risk of CO2 leakage and enhancing the long-term security of storage. This could make CCS a more viable and attractive option for industries looking to reduce their carbon footprint.
Shokrollahi emphasizes the importance of further research to expand the dataset and explore more complex systems. “Future studies should aim to enhance the predictive capabilities of our model by incorporating more diverse data and investigating alternative modelling techniques,” he says. “This will not only improve our understanding of CO2 dissolution in brine but also pave the way for more innovative and effective CCS solutions.”
The energy sector is at a critical juncture, and advancements like Shokrollahi’s XAI framework could be the key to unlocking the full potential of CCS. As the world continues to grapple with the challenges of climate change, such breakthroughs offer a glimmer of hope and a path forward towards a more sustainable future.
