Accelerating Carbon Capture and Storage Modeling using Fourier Neural Operators

Nested FNO for CO2 storage modeling

Nested FNO is a machine learning framework that provides unprecedented high-resolution high-fidelity 3D spatial-temporal predictions of CO2 gas saturation plume and pressure buildup. Nested FNO can handle a wide range of realistic reservoir and operational parameters and speeds up predictions nearly 700,000 times compared to state-of-the-art numerical simulators!

Prediction Accuracy

Nested FNO provides accurate gas saturation and pressure buildup prediction over 30 years of injection. The average gas saturation plume error is 1.8% and average pressure buildup error is 0.5%.

CCSNet.ai

Check out our web application ccsnet.ai that hosts the pre-train models. You can obtain real-time prediction with custom geological correlations, permeability statistic, reservoir depth, temperature, and dip angles. You can also use custom injection rate, injection location, and perforation design!

Citation

@article{wen2022accelerating,
  title={Accelerating Carbon Capture and Storage Modeling using Fourier Neural Operators},
  author={Wen, Gege and Li, Zongyi and Long, Qirui and Azizzadenesheli, Kamyar and Anandkumar, Anima and Benson, Sally M},
  journal={arXiv preprint arXiv:2210.17051},
  year={2022}
}