Future reservoir simulation workflows will to an increasing degree couple standard simulation models for multiphase flow with, for instance, models that represent complimentary physics (e.g. geomechanics, pore-scale models) or different spatial resolution (regional or near-well models). At the same time, today’s physics-based models will be supplemented with data-driven models that are fast but have poorly understood mathematical structures. The design of coupling strategies and solvers for such multi-model simulation tools is both highly challenging and critical to avoid simulation breakdown or results with poor accuracy.
2. Research goals
The goal of CSSR work package 5 is to device methodologies for ensuring the robustness and efficiency of multi-model simulation tools for subsurface engineering operations. Research will involve mathematical analysis of solvers for coupled problems, covering both physics-based models and novel data-driven models to be developed in work package 4. Implementation and testing of the coupling strategies on realistic problems will be a central activity.
The work package will produce a framework for robust coupling of multi-model simulations, implemented in open-source software that will be designed to facilitate easy coupling with individual simulation tools. The framework will exploit user insight into the character of individual models to enhance the coupling, and supplement this with on-the-fly tuning of the coupling strategy using machine learning techniques.
Department of Mathematics, University of Bergen
Coupling strategies and solvers for multi-model simulators
Solvers for coupled problems with automated parameter tuning
Robust model coupling for reservoir management
Tor Harald Sandve
Jan M. Nordbotten