A New Generation Framework for Petroleum
Reservoir Simulation
As part of the ACTI project, we have
developed a simulator framework IPARS (Integrated Parallel
Accurate Reservoir Simulator) to serve as a test-bed for multiphase
flow models, numerical discretizations, solvers, and upscaling.
IPARS
Users online help
Current Features
- portability between different platforms including PC under
DOS/Windows or Linux and PC clusters, RS6K and SP2, SGI,
Cray T3E and ... ;
- memory management for multiple fault blocks, multiblock
nonmatching grids and general geometry.
- multiple processors (ghost region updates and global reduce
operations);
- time step control;
- error management, cpu timing system;
- keyword input;
- visualization output for scalar and vector variables,
data tables etc.;
- a suite of linear solvers;
- well management tools for different types of injection
and production wells;
- multiple physical models including
- two phase (oil/water) implicit,
- two phase (oil/water) sequential (Fast Accurate),
- black oil (three phase, three component),
- two phase (air/water) implicit,
- single phase (slightly compressible) explicit,
- single phase (slightly compressible) implicit,
- EOS compositional (in review phase, also see EOS page),
- and more (some still in development phase) ...
- multiblock formulation using mortar spaces and implemented
with MACE ;
- multimodel formulation allowing for multinumerics and
multiphysics;
Recent Accomplishments
- We are now running Mega-size problems with the two-phase
and black oil models on our PC-clusters. The parallel scaling
of the code is very good.
- We have recently achieved multiphysics; coupling in which
IPARS models: black oil, two phase and single phase models
are coupled across interface. In the earlier phase of this
work, we completed multinumerics coupling.
- IPARS black oil model was succesfully used in the projects:
- Coupling of multiphase flow with geomechanics, in
collaboration with Sandia National Labs, see Coupled
Geomechanical Deformation and Reservoir Simulation ,
or here.
- Coupling of multiphase flow with single-well injectivity
modeling.
- Advances in parallel procedures for multiphase flow include
preconditioned GMRES with variable multistage preconditioners
(Uzawa, and IMPES family). These schemes have been tested
in the IPARS framework for two and three phase fully-implicit
models, and are the foundation of the excellent parallel
scalability of IPARS. Collaboration with Philippe Quandalle
(IFP) and Yuri Vassilevski (Russian Academy of Sciences)
was crucial to achieving the current level of performance.
- Collaboration with ICES (Institute for Computational Engineering
and Sciences) at UT has raised interesting possibilities
for applying recent computer science advances, such as coupling
the low level dynamic mesh data structures of the DAGH (Distributed
Adaptive Grid Hierarchy) package with the domain knowledge
framework embodied in IPARS, and using "space-filling
curves" for ordering the grid elements upon the processors.
Current features in detail:
- Multiple
fault blocks
- Multiblock
nonmaching grids
- Two phase (air/water) implicit physical model
- EOS
Web Page
- Muitiblock formulation
- Mortar spaces
- MACE
- Multinumerics formulation
- Multiphysics formulation
Recent accomplishments in detail:
- Mega-size problems
- Parallel scaling of the code
- Multiphysics coupling
- Multinumerics coupling
- Coupled Geomechanical Deformation and Reservoir Simulation
- Coupling of multiphase flow with single-well injectivity
modeling
Related Presentations
- Physical and computational domain decompositions for multiphysics,
multiphase, and multiscale porous media problems
- "Computational Challenges in Reservoir Simulation"
|
General geometry on
IPARS:
Dipping Reservoir
Multiblock
Small Angle Approximation
IPARS Collaborators:
- John Wheeler
- Mary Wheeler
- Krzysztof Banas
- Steve Bryant
- Srinivas Chippada
- Joe Eaton
- Carter Edwards
- Xiuli Gai
- Sebastien Lacroix
- Wonsuck Lee
- Qin Lu
- Bahareh Momken
- Myeong-Hwan Noh
- Manish Parashar
- Victor Parr
- Malgorzata Peszynska
- Philippe Quandalle
- Beatríce Rivière
- Kamy Sephernoori
- Yuri Vassilevski
- Peng Wang
- Ivan Yotov
|
