Multiscale methods, such as MsRSB initially proposed by SINTEF, have emerged to enhance runtime efficiency and pressure solve scaling in reservoir simulators using sequential splitting. Starting from the methods successfully implemented in the INTERSECT simulator for black-oil models, this project aims to investigate algorithmic enhancements essential for extending the technology to compositional simulation.
Conceptual illustration of an adaptive coarsening method for transport solvers, here with a hierarchy of three predefined meshes. Each coarse mesh is defined by aggregating blocks from the level below and defined by an integer partition vectors. Starting from a given partition vector in timestep n, the algorithm defines a cell-wise indicator function that measures whether each cell should be coarsened or refined to derive the new adaptive coarse mesh at timestep n+1
The project focuses on the following main activities:
Improve the overall solution strategy, including sequential fully implicit (SFI) methods, SFI methods with domain decomposition, and time-stepping methods.
Improved accuracy, efficiency, robustness, and scalability for the multiscale pressure solver. This includes work on multi-level and adaptive MsRSB formulations, potential parallelization on multi- and many-core architectures.
Improve accuracy, efficiency, robustness, and scalability of the transport solves. This includes topics such as inexact Newton solvers, local time-stepping, localization of Newton updates, adaptive coarsening, methods for optimal ordering, and use of high-resolution spatial discretizations.
Conduct a series of numerical experiments on simplified models as well as simulation models representative of real assets to build a strong case for industrialization.
Key Factors
Project duration
01/01/2018 - 31/12/2021
References
Ø. Klemetsdal, A. Moncorgé, O. Moyner, and K.-A. Lie. “Additive Scharz preconditioned exact Newton method as a nonlinear preconditioner for multiphase porous media flow”. In: ECMOR XVII–17th European Conference on the Mathematics of Oil Recovery, Online, 14–17 September, 2020. European Association of Geoscientists & Engineers, 2020. DOI: 10.3997/2214-4609.202035050.
Ø. S. Klemetsdal, A. Moncorgé, H. M. Nilsen, O. Møyner, and K.-A. Lie. “An adaptive sequential fully implicit domain-decomposition solver”. In: SPE Journal (Sept. 2021), pp. 1–13. DOI: 10.2118/ 203991-PA.
Ø. S. Klemetsdal, O. Møyner, A. Moncorgé, H. M. Nilsen, and K.-A. Lie. “High-resolution compositional reservoir simulation with dynamic coarsening and local timestepping for unstructured grids”. In: SPE Journal 26.06 (Sept. 2021), pp. 4157–4173. doi: 10.2118/203982-PA.
Ø. Klemetsdal, A. Moncorgé, O. Møyner, and K.-A. Lie. “A numerical study of the additive Schwarz preconditioned exact Newton method (ASPEN) as a nonlinear preconditioner for immiscible and compositional porous media flow”. In: Computational Geosciences (Sept. 2021). DOI: 10.1007/s10596-021-10090-x.
G. Linga, O. Møyner, H. M. Nilsen, A. Moncorgé, and K.-A. Lie. “An implicit local time-stepping method based on cell reordering for multiphase flow in porous media”. In:Journal of Computational Physics: X 6 (Mar. 2020), p. 100051. DOI: 10.1016/j.jcpx.2020.100051.ad.
H. M. Nilsen, A. Moncorge, K. Bao, O. Møyner, K.-A. Lie, and A. Brodtkorb. “Comparison between algebraic multigrid and multilevel multiscale methods for reservoir simulation”. In: ECMOR XVII– 17th European Conference on the Mathematics of Oil Recovery, Online, 14–17 September, 2020. European Association of Geoscientists & Engineers, 2020. DOI: 10.3997/2214-4609.202035063.
