Scalable Simulation of Subsurface CO2 Sequestration Processes: the GEOS Open-Source Platform
Hamdi Tchelepi, Stanford
Video Recording
Slides (pptx, pdf)
Abstract:
Subsurface CO2 sequestration at the scale of Gigatons per year will be necessary to meet the U.N. targets of reducing global CO2 emissions. The subsurface porous formations include deep saline aquifers and depleted hydrocarbon reservoirs. We describe a numerical simulation platform (GEOS) of the physics that governs the dynamics of fluid flow and fluid-solid interactions. The talk includes a statement of the governing equations, numerical discretization methods, and scalable nonlinear and linear solvers of coupled systems of equations in heterogeneous porous formations. The architecture of the open-source HPC platform is (GEOS) is described. Large-scale use-cases will be presented, and simulation results of the injection and post-injection periods will be discussed.
Bio:
Research: Numerical simulation of flow, transport, and fluid-structure interactions in multiscale porous media.
Areas of ongoing activity: (1) modeling and simulation of unstable miscible and immiscible fluid flow in heterogeneous porous media, (2) development of multiscale numerical solution algorithms for coupled mechanics and multiphase fluid flow in large-scale subsurface formations, and (3) development of stochastic numerical methods that quantify the uncertainty associated with predictions of nonlinear fluid-structure dynamics in heterogeneous porous media.
The application areas include reservoir simulation and subsurface CO2 sequestration at scale. An area of growing interest is modeling and high-fidelity numerical simulation of species transport and fluid-structure interactions in the next-generation of Lithium-ion batteries.
Teaching: I teach courses on multiphase flow in porous media and numerical reservoir simulation.
Professional Activities: President's Individual Achievement Award, sponsored by Chevron and Schlumberger, for successful completion of the Intersect Project (next generation reservoir simulator), 2003; Co-Director, Stanford Reservoir Simulation Affiliates Program (SUPRI-B), 2006-present; Editorial board, Transport in Porous Media, 2005-2010; advisory panel, Center for Computational Earth and Environmental Science, 2005-present; graduate admissions committee, Department of Energy Resources Engineering, 2004-2017; Editorial board, SPE Journal, 2000-present; member, SPE, AGU, APS, SIAM, 1999-present; Edmund W. Littlefield Fellow, 1993-94
Summary:
Motivation: CO2 sequestration in subsurface formations is a key technology for mitigating climate change.
MALSTROM project:https://www.llnl.gov/article/46901/llnl-partners-open-access-co2-storage-simulator
DOE Report: Accelerating Breakthrough Innovation in Carbon Capture, Utilization, and Storage
There are many projects that capture CO2 from point sources or directly from the atmosphere and need subsurface storage
Simulation of CO2 sequestration
Model the process of injecting CO2 deep underground and the long post-injection evolution of the CO2 plume(s)
Supercritical state: High density, low viscosity
Consideration:
Safety
Injectivity
Capacity
Containment
Interactions
Long-term fate
Phenomena to model
Multiphase flow dynamics
Hydro-fracturing
Hydro-shearing
Fault reactivation
Induced seismicity
Well integrity
Surface deformation
Challenges
Multi-physics
Multi-scale
HPC
Sequestration process
Capture CO2 from either point or distributed sources. Point sources include cement plants, steel manufacturing, fertilizer production, biomass energy, natural gas power plants, blue hydrogen production, etc. There is recent interest as well in “direct air capture,” pulling CO2 directly from the atmosphere. This is energy inefficient compared to point source capture but may be necessary long-term to achieve climate change targets.
Compress the CO2, and inject under ground (800 meters +) into small pores within rock formations
Salt-water aquifer
Supercritical CO2 initially exists as a separate phase, but begins dissolving into surrounding in-situ brine.
Trapping process:
Structural: Solid, impermeable rock layers above injection formations to keep the CO2 contained.
Residual: CO2 trapped in micron-scale pores within the rock
Dissolution trapping: CO2 dissolves into surrounding salt water and no longer buoyant
Mineral trapping: CO2 reacts with surrounding rock, creating new solid minerals (e.g. Ankerite: https://www.mindat.org/min-239.html)
Mineral trapping is the biggest hope but it takes a long time, so main focus of project design is to rely on structural, residual and dissolution trapping
CO2 sequestration dynamics are challenging to model because several important effects are at play: gravity currents, convective dissolution, geomechanical interactions
Model:
Darcy’s Law-based model tuned to the specific dynamics of CO2 (https://www.ldeo.columbia.edu/~martins/climate_water/lectures/darcy.html)
Fluid dynamics of CO2-saturated water falling into CO2-free in-situ water as “finger” shapes, with low-Reynolds number
Length scales:
Rock pores are micron-size
CO2 fingers start at the cm scale
Rock formations are kms in size
Time scales: years - Kyears
10's to 100's years for dissolution dynamics to fully engage
100s of years for dissolution process to complete
Data-limited:
Cores and wellbore logging tools at wellbore locations
Well-based observations of flow rates, pressure, temperature, fluid composition
Remote geophysical imaging methods. Mainly active 3D/4D seismic imaging and passive microseismic monitoring and tomography. Other techniques such as satellite-based InSAR, electromagnetics, gravity surveys, and various types of fiber optic monitoring may be applicable depending on site characteristics.
Data assimilation and uncertainty quantification are critical for aligning model to reality
Multiscale poromechanics
Flow of CO2 fluid through solid rock matrix
Structural mechanics of the rock matrix to evaluate its stability and impact of CO2 injection on the rock matrix
Fluid-structural interactions:
Multiple fluids: supercritical CO2, salt-water, possibly hydrocarbons (multiple fluid phases)
Multi-phase flow
Interphase mass transfer
Flow-GeoMechanics Simulation: Physics, multi-scale, coupled flow+mechanics, IHU discretization, non-linear solvers, scalable algorithms
Simulator objectives:
Unstructured geo grids: faults, fractures
Tightly coupled flow, transport, mechanics
Solvers: algebraic and multi-level (different solvers for each scale/physics)
Scalable HPC
Validation
Related communities to be inspired from:
Oil and gas simulation
Subsurface hydrology
Careful use of available data
Understanding of structural and regional geology
Seismic information
Need to constantly adapt the model based on changes in data: couple forward and inverse problems
Confidence and processes improve as you keep doing projects
Model uncertainty
Physics at pore-scale is well-known and we are confident about our approximations of this process
Most of the uncertainty is in the details of the geology of the target formation and the layers around it
Seismic surveys give you resolution at many meters
Need to create fine-grained of geology with appropriate discontinuities and roughness that is consistent with the seismic survey data
Use-case for generative models (e.g. ML) that create rough geology models
HPC scalability
GEOS can be used at high resolution on leading-scale supercomputers
Or at lower scale on smaller clusters
Goal: suites of multi-fidelity resolutions, coupled with neural surrogates