Energy minimizing bases for efficient multiscale modeling and linear solvers in reservoir simulation

We demonstrate the applicability of energy minimizing bases for two-phase flow simulations, and we highlight several advantageous properties that they possess. We show how they can be implemented to obtain efficient serial (algebraic multigrid) and parallel (Additive Schwarz with coarse space correction) linear solvers for large-scale heterogeneous problems. We also confirm with experiments that using such bases for coarsening a problem produces a numerical solution of quality comparable to that of other available multiscale techniques. Finally, we also indicate how different coarsening levels can be used in different regions, in order to keep a finer resolution near certain areas of interest and significantly coarser resolution elsewhere. A multilevel approach is adopted to efficiently construct the basis functions for the coarser resolution.