 |
| Titel |
Integrating surrogate models into subsurface simulation framework allows computation of complex reactive transport scenarios |
| VerfasserIn |
Marco De Lucia, Thomas Kempka, Janis Jatnieks, Michael Kühn |
| Konferenz |
EGU General Assembly 2017
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 19 (2017) |
| Datensatznummer |
250154127
|
| Publikation (Nr.) |
 EGU/EGU2017-19187.pdf |
|
|
|
|
|
| Zusammenfassung |
| Reactive transport simulations - where geochemical reactions are coupled with hydrodynamic
transport of reactants - are extremely time consuming and suffer from significant numerical
issues. Given the high uncertainties inherently associated with the geochemical
models, which also constitute the major computational bottleneck, such requirements
may seem inappropriate and probably constitute the main limitation for their wide
application.
A promising way to ease and speed-up such coupled simulations is achievable employing
statistical surrogates instead of "full-physics" geochemical models [1]. Data-driven surrogates
are reduced models obtained on a set of pre-calculated "full physics" simulations,
capturing their principal features while being extremely fast to compute. Model
reduction of course comes at price of a precision loss; however, this appears justified
in presence of large uncertainties regarding the parametrization of geochemical
processes.
This contribution illustrates the integration of surrogates into the flexible simulation
framework currently being developed by the authors’ research group [2]. The high level
language of choice for obtaining and dealing with surrogate models is R, which profits from
state-of-the-art methods for statistical analysis of large simulations ensembles. A stand-alone
advective mass transport module was furthermore developed in order to add such capability
to any multiphase finite volume hydrodynamic simulator within the simulation
framework.
We present 2D and 3D case studies benchmarking the performance of surrogates and "full
physics" chemistry in scenarios pertaining the assessment of geological subsurface
utilization.
[1] Jatnieks, J., De Lucia, M., Dransch, D., Sips, M.: "Data-driven surrogate model
approach for improving the performance of reactive transport simulations.", Energy Procedia
97, 2016, p. 447-453.
[2] Kempka, T., Nakaten, B., De Lucia, M., Nakaten, N., Otto, C., Pohl, M., Chabab
[Tillner], E., Kühn, M.: "Flexible Simulation Framework to Couple Processes in Complex 3D
Models for Subsurface Utilization Assessment.", Energy Procedia, 97, 2016 p.
494-501. |
|
|
|
|
|
|