Background
Multiscale technology has been researched since around the turn of the century. Many academic studies have indicated that the methods have the potential to provide fast computation of approximate, but representative flow solutions that can be used to verify parameter and model choices early in the reservoir charact erization process. However, this is yet to be demonstrated in a full industry-relevant setting.
To mature multiscale simulation to an industrial reality, the project will address the following research challenges:
- Develop multiscale methods that are flexible with respect to the size and shape of coarse and fine grid blocks. Develop suitably prescribed boundary conditions on interfaces of irregularly-shaped coarse blocks, coarse-grid operators that produce high quality coarse-scale solutions, as well as fast converging iterative schemes for coupled pressure and transport.
- Construct algorithms for automatic coarsen ing of complex models that adapt to important features such as wells, short-range and long-range geological heterogeneity, and faults.
- Investigate concurrency of multiscale methods and develop prototype implementations that fully utilize multi-core an d many-core hardware to scale well on current and future hardware architectures.
- Provide proof-of-concept prototypes of workflows that utilize variable-fidelity multiscale simulation to improve reservoir characterization, uncertainty quantification, a nd production optimization.
Project results
SINTEF had proviously proposed the so-called multiscale restriction-smoothed basis (MsRSB) method that seemed promising to overcome fundamental problems with multiscale methods that prevented their industrialization. In the project, we proved that MsRSB can overcome these fundamental problems and be implemented in a commercial setting.
The method is still state-of-the art, has been proven in an operational environment by SLB, and is today the main engine of their INTERSECT multiscale sequential fully implicit simulator, offering enhanced simulation performance on heterogeneous, high-resolution geomodels, e.g., for use in geo-screening.
Work conducted in this and proceeding projects was quoted when Møyner received the SIAM Geosciences Early Career award in 2019 and Lie was elected SIAM Fellow in 2020.