David Trebotich and Carl Steefel (left to right) stand in front of Cori NERSC’s supercomputer system.
“An Exascale Subsurface Simulator of Coupled Flow, Transport, Reactions and Mechanics” is one of two Lawrence Berkeley National Laboratory (Berkeley Lab), Earth and Environmental Sciences Area-led projects recently selected by the Department of Energy (DOE) Exascale Computing Project (ECP) to be funded at $10M over the next four years.
Berkeley Lab Principal Investigator, Carl Steefel (senior scientist and Geochemistry Department Head for the Earth & Environmental Sciences Area), with David Trebotich (staff scientist, Applied Math Department, Computational Research Division) serving as deputy, leads a multi-laboratory team, that includes Lawrence Livermore National Laboratory (LLNL) and the National Energy Technology Laboratory, to develop an exascale subsurface simulator that will provide the ability to predict the evolution of wellbores and fracture networks, across scales from small pores in the rocks to large underground reservoirs.
The three colored regions depict a relevant subsurface spatial scale, span roughly 4 orders of magnitude, and require an exascale machine. Red region is pore scale resolution (100 nm on 1 cm domain); Orange region covers the intermediate scale in the vicinity of key features like fractures and wells (1 mm on 1 m domain); Green region is an ultra-high resolution reservoir-scale model (1 m on 1 km domain).
Subsurface geologic structures can be exploited for enhanced energy extraction and storage, but to do so requires a sound understanding of and predictive capability for the coupled thermal, hydrological, chemical, and mechanical (THCM) processes that control the success or failure of energy-related endeavors, including geologic CO2 sequestration, petroleum extraction, geothermal energy and nuclear waste isolation. The inherent multiscale nature of the subsurface, however, makes predictions of these subsurface processes difficult, particularly when relatively small-scale features like fractures or damage zones around wellbores can disproportionately affect the larger-scale system behavior.
Current petascale (1 quadrillion, or 1015, floating point operations per second) computer architectures and the software that executes on them cannot provide the computing power needed to solve the subsurface multi-scale problem. The DOE ECP is advancing the development of exascale supercomputing architectures capable of performing a billion billion, or 1018, floating point operations per second, as well as adapting existing applications that can take advantage of the new architectures.
To advance predictive understanding of the multiscale nature of the subsurface, this project will couple two mature petascale code bases: 1) Chombo-Crunch (developed at Berkeley Lab), which models subsurface flow at pore and continuum scales coupled to multicomponent geochemistry, and 2) the GEOS code (developed at LLNL), which models geomechanical deformation and fracture+Darcy flow at a variety of scales. Adapted for exascale systems, the applications will be able to simulate the behavior of subsurface flows across both spatial and time scales.
Results of high resolution pore scale simulations of topology of a single fracture in dolomite using Chombo-Crunch. Top: Initial geometry of fracture. Bottom: Modified geometry after 18 hours showing chemical erosion of fracture “pillars”.
Learn more about the Project and the ECP by going to the Berkeley Lab’s News Center.
Meet the Team Members
- Lawrence Berkeley National Laboratory—Carl Steefel, David Trebotich, Brian Van Straalen
- Lawrence Livermore National Laboratory—Randolph Settgast, Joseph Morris, Susan Carroll
- National Energy Technology Laboratory—Grant Bromhal