• Traditional physics based simulation approaches for inverse modeling and forecasting in geologic CO2 sequestration (GCS), is a very time consuming process. In this work, SMART researchers have developed a deep learning assisted workflow to speed up this process by more than 90 minutes to only 2.7 seconds (~2000x) in one forward prediction run in the inverse modeling process.
  • Key aspects of this approach are:
    • A deep learning model to predict the pressure/saturation evolution in large-scale storage reservoir.
    • A feature coarsening technique to extract the most representative information and perform the training and prediction at the coarse scale, and further recover the resolution at the fine scale by 2D piecewise cubic interpolation.
    • Applying this technique as forward model in the inverse modeling process where a classical data assimilation approach, ES-MDA-GEO, is used.
  • This workflow is demonstrated with a reservoir model (~1.34 million grid cells) built upon Clastic Shelf Model for history matching and forecasting under uncertainty.
  • More details about feature coarsening based deep learning model can be found in the preprint: https://arxiv.org/abs/2105.03752; Details about leveraging ES-MDA-GEO for inverse modeling in GCS can be found: https://www.sciencedirect.com/science/article/pii/S1750583619304062; POC for this work: Bailian Chen (bailianchen@lanl.gov).