Numerical simulation of subsurface flow and transport is a mature technology, and has been extensively used to address technical challenges in subsurface science and engineering. Over the decades, several well-known simulators have been developed, such as PFLOWTRAN, OpenGeoSys, TOUGH, STOMP, and FEHM. All of these simulators are developed for research purposes and can adequately represent the complex physical and chemical processes of flow and transport in porous and fractured media. However, there is still significant room for improvement in computing efficiency and accuracy, when comparing with recent developments in oil/gas reservoir simulations. In this study, we reviewed the most well-known oil/gas reservoir simulators along with other subsurface flow-transport simulators, investigated modern modeling technology, and developed a general modeling approach. The approach benefits from the integration of advanced modeling techniques, such as Automatic Differentiation (AD) for calculating the Jacobian matrix, Adaptive Implicit Methods (AIM) for solving non-linear equations, Continuation-Newton Algorithm (CN) for improving time-step sizes, MultiPoint Flux Approximations (MPFA) for flux calculation, and better preconditioners and linear solvers (such as the proved efficient preconditioners used in oil/gas reservoir simulations: CPR, AMG, FASP) for solving linear equations. A distributed modeling approach was developed for the implementation of extremely large-scale simulations (with 0.1 to 1 billion gridblocks), which makes pre- and post-processing for these simulations very easy. In addition, we developed a parallel-computing technique for hybrids of distributed, memory-shared architectures and mixed-computing platforms, such as the combinations of many-core, multi-core processors and GPU-based accelerators. Based on this study, an advanced general-purpose simulator of subsurface multiphase flow and transport was developed. The generality of the new simulator roots from its flexibility in the calculation of fluxes for Darcy, Non-Darcy, Newtonian, Non-Newtonian fluids or any undefined fluids, for which their flux can be determined through multivariable interpolation. Testing examples show the new simulator can easily achieve one order of magnitude speed-up, compared to existing similar simulators on the same computing platform.
In this talk, Dr. Zhang will also discuss pros and cons of the TOUGH family codes. Suggestions for future improvement of the TOUGH codes will be given.
About the Speaker: Keni Zhang
Dr. Keni Zhang is a consultant for subsurface flow-transport simulations and development of flow-transport simulators in the fields of geological CO2 sequestration, nuclear waste disposal, geothermal engineering, and oil/gas reservoir development. He worked at LBNL as a postdoc scholar and geological scientist during 2000-2017 and was the primary developer of the TOUGH2-MP codes. He was a full professor at Beijing Normal University during 2009-2015.
Host: Quanlin Zhou