Deep geothermal targets are generally located within complex geological systems, such as multi-scale fault zones, generally characterized by a strong spatial variability of many of its spatially distributed properties among which permeability, porosity, compressibility, thermal conductivity, volumetric heat capacity.
The development of geothermal energy generation is closely linked to thermal and hydrogeological knowledge of the subsurface aquifers. Numerical modeling here appears as a tool to delineate development risks induced by limited geological data at great depths.
Computational times of deep geothermal models can be rather long (tens of minutes to hours). Considering a given deterministic model, what are are the most efficient numerical methods to quantify uncertainties related to subsurface physical properties on reservoir production forecasts ?
Hi Morgan,
DIANA FEA is used for that. spatial and or probabilistic parameters combined with correct damage material models.
But off course, normally they are large models that take time that is completely true.
Thank you for your input.
More than a numerical code, I am interested in numerical methods to quantify uncertainties on physical parameters that are homogeneous within an hydrogeological unit.
Speaking of numerical code, I use PumaFlow (reservoir simulator) and CougarFlow (uncertainty management & assisted history matching) from Beicip-Franlab.
Do you mean ?
M AHON, W. (1977). Chemistry and Geothermal Systems .
Best regards.
Morgan yes.
Please also refer Koppel Mayer and Haenel R 1974 Geothermic with ......monograph series, 1-4, Geo publications Association, New York .
Hi Morgan,
Do you know Pest?
Model-Independent Parameter Estimation and Uncertainty Analysis
http://www.pesthomepage.org
We have used it for both hydrogeological and geothermal models. As one example we quantified uncertainties of predicted temperatures according to the uncertainty of the spatial input of rock thermal properties.
However, I can give you no assessment how this tool performs in terms of efficiency compared to other methods.
Hi Sven,
I used Pest when working on FELFOW with integrated package FePest. It is indeed a perfect tool for such uncertainty definition.
However, most numerical methods propose in Pest require a large number of runs to be validated. That we cannot afford when dealing with numerical models that take hours for a run.
Did you write an article on the utilization of Pest in geothermal field ?
Best regards.
I have used FePest to calibrate the spatial variable thermal rock properties of a non-stationary 3D hydraulic-thermal model of the region of Denmark. The model contains several million nodes and the calibration took 1week+ using 24 cores. So from my experience I totally agree that using Pest potentially demands for a larger amount of computation cores. The described study is not published yet, but we have used the same approach in a similar study of the Glückstadt-Graben (2 parts)
https://www.researchgate.net/publication/301888849_Improving_the_temperature_predictions_of_subsurface_thermal_models_by_using_high-quality_input_data_Part_1_Uncertainty_analysis_of_the_thermal-conductivity_parameterization
https://www.researchgate.net/publication/301743521_Improving_the_temperature_predictions_of_subsurface_thermal_models_by_using_high-quality_input_data_Part_2_A_case_study_from_the_Danish-German_border_region
A recent study that focuses on the uncertainty business using fepest was provided by Søren Erbs Poulsen
for example.
https://www.researchgate.net/publication/319236210_Inverse_geothermal_modelling_applied_to_Danish_sedimentary_basins
Regards
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis