Review of progress and implication of machine learning in geological carbon dioxide storage
Mahlon Kida Marvin, Victor Inumidun Fagorite, Alhaji Shehu Grema, Mohammed Dahiru Aminu, Aliyu Buba Ngulde, Zakiyyu Muhammad Sarkinbaka
Abstract
Deep underground rock formations are widely used for geological carbon dioxide (CO2) storage due to their large-scale, long-term capacity. However, geophysical and petrophysical complexities can lead to challenges such as gas migration and potential leaks, posing risks to groundwater and subsurface systems. Recent advancements increasingly integrate Artificial Intelligence (AI) and Machine Learning (ML) to mitigate these risks and enhance CO2 storage efficiency. This review explores ML applications in geological CO2 storage, highlighting recent advancements and their implications. ML has demonstrated effectiveness in enhancing CO2 storage efficiency. However, the complexities of geological storage necessitate further improvements in ML model applicability, particularly in real-world projects. Since ML models depend on the availability of data, ensuring high data integrity and quality is crucial. Moreover, CO2 storage projects involve significant risks and uncertainties, making advanced probabilistic ML models essential for quantifying uncertainties and mitigating associated risks. Lastly, integrating real-time monitoring systems with sensor data and ML algorithms can enhance anomaly detection, provide early warnings, and enable timely interventions. Addressing these challenges will strengthen the adoption of advanced ML techniques in geological CO2 storage, improving efficiency, safety, and reliability.