Trending Content

Deep Learning for Rapid Geomodelling and Carbon Storage Forecasting

Add to Cart
Course Credit: 0.15 CEU, 1.5 PDH

Subsurface earth models (referred as geomodels) are crucial for characterizing complex subsurface systems. A deep-learning-based generative method is developed as an alternative to traditional geomodel generation procedure. The generative method comprises two deep-learning techniques based on advanced autoencoder followed by autoregressor. In the conditional geomodel generation, the generative workflow can rapidly generate an ensemble of thousands of geomodels similar to an user-defined source geomodel in seconds, which ultimately facilitates the control and manipulation of the generated geomodels.

Deep learning along with transfer learning can be used to rapidly forecast the spatiotemporal evolutions of saturation and pressure in large, heterogeneous reservoir. The deep-learning-based rapid forecasting enables probabilistic assessment of the 3D plume volume evolving during the injection and post injection stages for the geological carbon storage (GCS). The deep learning capabilities are demonstrated on the field-scale SACROC geomodel with permeability-porosity heterogeneity. The new workflow reduces the computational cost and time of reservoir simulation, under variable injector-producer locations and variable injection rates. This is essential for large-scale probabilistic assessment of carbon storage and containment with possibility of helping low-cost adaptive monitoring for the carbon storage plume over hundreds of years. All content contained within this webinar is copyrighted by Siddharth Misra and its use and/or reproduction outside the portal requires express permission from Siddharth Misra.

Post Tags

 1 chapter
  1 downloadable resource

Course Chapters

  • 1Deep Learning for Rapid Geomodelling and Carbon Storage Forecasting - Chapter 1
    Media Type: Video


Earn credits by completing this course0.15 CEU credit1.5 PDH credits


Siddharth Misra
Yusuf Falola