Scalable Algorithms for Tightly-Coupled Hydro-mechanical Modeling of Geologic CO2 Sequestration

Modeling the injection and long-term storage potential of carbon dioxide in saline aquifers involves an intricate coupling between fluid flow and geomechanical behavior. In particular, in faulted and fractured reservoirs there exists a tight feedback between pressure, concentration, permeability, stress, and solid constitutive behavior. From a modeling standpoint, it is therefore often necessary to solve the governing PDEs for all unknown fields simultaneously, rather than using a staggered or loosely-coupled scheme. Here, we describe our recent work developing scalable Newton-Krylov solvers to efficiently model large-scale CO2 injection operations with multiphysics fidelity. To motivate the work, we discuss an ongoing study in which detailed forward models of CO2 injection are used to interpret InSAR observations of surface uplift due to pressurization of the storage interval. Resolving the necessary spatial scales leads to very large algebraic problems that must be solved in each Newton iteration. In our solver methodology, we focus on the use of block-structured preconditioners to deal with the inherent ill-conditioning of these systems. We discuss effective choices for Schur-complement approximations and sub-block preconditioners. An algebraic multigrid variant of the proposed methodology leads to mesh-independent Krylov convergence and good weak-scalability. This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344.