Effects of transient processes for thermal simulations of the Central European Basin

Degen, Denise; Cacace, Mauro

doi:10.5194/gmd-14-1699-2021

Cited by 3 publications

(4 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This can be further highlighted by looking at the intrusive counterpart of the method. Degen and Cacace (2021) show, that the results of the sensitivity analysis are not impacted by the model accuracy, which demonstrates that the general characteristic is the same for all models.…”

Section: Requirement 4: Robustnessmentioning

confidence: 57%

“…The non-intrusive RB method maintains the input-output relationship, in contrast to, for instance, PINNs. As shown for its intrusive counterpart, the results of a sensitivity analysis do not change significantly with the surrogate model accuracy (Degen and Cacace, 2021), demonstrating that the surrogate does not change the general characteristics of the original problem and making it a suitable technique to ensure the feasibility of global sensitivity studies also for large-scale models.…”

Section: Challenge 1: Sensitivity Analysismentioning

confidence: 91%

“…On the other hand, the reduced model requires 3 ms for But even if we want to solve for both variables for all time steps, we are nearly three orders of magnitude faster with the reduced model. From previous studies using the intrusive RB method, this speed-up is expected to be higher for real case models because the complexity of these models increases slower than the degrees of freedom (Degen et al, 2020a;Degen and Cacace, 2021;.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Perspectives of Physics-Based Machine Learning for Geoscientific Applications Governed by Partial Differential Equations

Degen

Caviedes‐Voullième²,

Buiter

et al. 2023

Preprint

View full text Add to dashboard Cite

Abstract. An accurate assessment of the physical states of the Earth system is an essential component of many scientific, societal and economical considerations. These assessments are becoming an increasingly challenging computational task since we aim to resolve models with high resolutions in space and time, to consider complex coupled partial differential equations, and to estimate uncertainties, which often requires many realizations. Machine learning methods are becoming a very popular method for the construction of surrogate 5 models to address these computational issues. However, they also face major challenges in producing explainable, scalable, interpretable and robust models. In this manuscript, we evaluate the perspectives of geoscience applications of physics-based machine learning, which combines physics-based and data-driven methods to overcome the limitations of each approach taken alone. Through three designated examples (from the fields of geothermal energy, geodynamics, and hydrology), we show that the non-intrusive reduced basis method as a physics-based machine learning approach is able to 10 produce highly precise surrogate models that are explainable, scalable, interpretable, and robust.

show abstract

Section: Requirement 4: Robustnessmentioning

confidence: 57%

Section: Challenge 1: Sensitivity Analysismentioning

confidence: 91%

mentioning

confidence: 99%

See 1 more Smart Citation

Perspectives of Physics-Based Machine Learning for Geoscientific Applications Governed by Partial Differential Equations

Degen

Caviedes‐Voullième²,

Buiter

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Therefore, we could, for instance, employ a Robin boundary condition to allow an interaction between the subsurface and the atmosphere. A previous study 38 has shown that a variation of the value of the boundary condition does not majorly impact the sensitivities of the model response as long as we consider a Dirichlet boundary condition. This is the reason because the value of the boundary condition is the same in every realization.…”

Section: Uncertainty Quantification Mapsmentioning

confidence: 94%

Uncertainty quantification for basin-scale geothermal conduction models

Degen

Veroy

Wellmann

2022

Sci Rep

Self Cite

View full text Add to dashboard Cite

Geothermal energy plays an important role in the energy transition by providing a renewable energy source with a low CO2 footprint. For this reason, this paper uses state-of-the-art simulations for geothermal applications, enabling predictions for a responsible usage of this earth’s resource. Especially in complex simulations, it is still common practice to provide a single deterministic outcome although it is widely recognized that the characterization of the subsurface is associated with partly high uncertainties. Therefore, often a probabilistic approach would be preferable, as a way to quantify and communicate uncertainties, but is infeasible due to long simulation times. We present here a method to generate full state predictions based on a reduced basis method that significantly reduces simulation time, thus enabling studies that require a large number of simulations, such as probabilistic simulations and inverse approaches. We implemented this approach in an existing simulation framework and showcase the application in a geothermal study, where we generate 2D and 3D predictive uncertainty maps. These maps allow a detailed model insight, identifying regions with both high temperatures and low uncertainties. Due to the flexible implementation, the methods are transferable to other geophysical simulations, where both the state and the uncertainty are important.

show abstract

Perspectives of physics-based machine learning strategies for geoscientific applications governed by partial differential equations

Degen,

Caviedes Voullième,

Buiter

et al. 2023

Geosci. Model Dev.

View full text Add to dashboard Cite

Abstract. An accurate assessment of the physical states of the Earth system is an essential component of many scientific, societal, and economical considerations. These assessments are becoming an increasingly challenging computational task since we aim to resolve models with high resolutions in space and time, to consider complex coupled partial differential equations, and to estimate uncertainties, which often requires many realizations. Machine learning methods are becoming a very popular method for the construction of surrogate models to address these computational issues. However, they also face major challenges in producing explainable, scalable, interpretable, and robust models. In this paper, we evaluate the perspectives of geoscience applications of physics-based machine learning, which combines physics-based and data-driven methods to overcome the limitations of each approach taken alone. Through three designated examples (from the fields of geothermal energy, geodynamics, and hydrology), we show that the non-intrusive reduced-basis method as a physics-based machine learning approach is able to produce highly precise surrogate models that are explainable, scalable, interpretable, and robust.

show abstract

Effects of transient processes for thermal simulations of the Central European Basin

Cited by 3 publications

References 48 publications

Perspectives of Physics-Based Machine Learning for Geoscientific Applications Governed by Partial Differential Equations

Perspectives of Physics-Based Machine Learning for Geoscientific Applications Governed by Partial Differential Equations

Uncertainty quantification for basin-scale geothermal conduction models

Perspectives of physics-based machine learning strategies for geoscientific applications governed by partial differential equations

Contact Info

Product

Resources

About