Abstract:
Numerical simulations of subsurface fluid and heat flow are commonly controlled manually via input files or from graphical user interfaces (GUIs). Manual editing of input files is often tedious and error-prone, while GUIs typically limit the full capability of the simulator. Neither approach lends itself to automation, which is desirable for more complex simulations. We propose an alternative approach based on the use of scripting. To this end we have developed Python libraries for scripting subsurface simulations using the SHEMAT and TOUGH2 simulators. For many problems the entire modeling process including grid generation, model setup, execution, post-processing and analysis of results can be carried out from a single Python script. Through example problems we demonstrate some of the potential power of the scripting approach, which does not only make model setup simpler and less error-prone, but also facilitates more complex simulations involving, for example, multiple model runs with varying parameters (e.g. permeabilities, heat inputs, and the level of grid refinement). It is also possible to apply the developed methods for extending the functionality of graphical user interfaces. Basing our approach on the Python language makes it simple to take advantage of other libraries available for scientific computation, with sophisticated analysis of results often a matter of a single function call. We envisage many other possible applications of the approach, including linking with geological modeling software, running stochastic ensembles of models and hybrid modeling using multiple interacting simulators.
