Reservoir Simulation and Modeling with Deep Learning

University of Calgary and IBM TJ Watson Research


  • Dr. John Chen, Engineering, University of Calgary
  • IBM Thomas J. Watson Research Center, Cambridge, Massachusetts
  • Industry partners

The objective of this research project is to investigate the development of machining learning in reservoir simulations. The outcome will advance algorithms, history matching and optimization techniques used in the simulations. The development of a novel reservoir simulator that can efficiently read a model and accurately adjust the parameters based on the historical data and the current changing data will be emphasized. The evidence of success will be visible when new models are validated against history matching studies. History matching will be advanced in fast, quality validations of comprehensive robust models. Optimization using the new algorithms based on multiple geological realizations is the methodology that will be used to bring together the current and historical data from specific to general reservoirs. These algorithms reduce simulation cost and provide the required accuracy and the best process to use in optimization, in order to reduce fiscal costs and increase the clean recovery of hydrocarbons.