Deep?Learning?Based Inverse Modeling Approaches: A Subsurface Flow Example

نویسندگان

چکیده

Deep-learning has achieved good performance and demonstrated great potential for solving forward inverse problems. In this work, two categories of innovative deep-learning-based modeling methods are proposed compared. The first category is deep-learning surrogate-based inversion methods, in which the Theory-guided Neural Network (TgNN) constructed as a surrogate problems with uncertain model parameters. By incorporating physical laws other constraints, TgNN can be limited simulation runs accelerate process significantly. Three proposed, including gradient method, Iterative Ensemble Smoother training method. second direct-deep-learning-inversion constrained geostatistical information, named TgNN-geo, framework direct modeling. neural networks introduced to approximate random parameters solution, respectively. order honor prior information parameters, network approximating trained by using observed or generated realizations. Then, minimizing loss function estimation approximation solution simultaneously obtained. Since incorporated, direct-inversion method based on TgNN-geo works well, even cases sparse spatial measurements imprecise statistics. Although general nature, thus applicable wide variety problems, they tested several subsurface flow It found that satisfactory results obtained high efficiency. Moreover, both advantages disadvantages further analyzed methods.

برای دانلود باید عضویت طلایی داشته باشید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Prediction-Focused Subsurface Modeling: Investigating the Need for Accuracy in Flow-Based Inverse Modeling

The objective of most formulations of inverse modeling in the Earth Sciences is to estimate the model parameters given the observation data. In this paper, an additional element to these formulations is considered, namely, the prediction for which the models are built. An example of such modeling is the prediction of solute transport using geological models of the subsurface constrained to exis...

متن کامل

Improved inverse modeling for flow and transport in subsurface media: Combined parameter and state estimation

[1] Current approaches for inverse modeling (IM) to estimate flow and transport properties in subsurface media implicitly assume that uncertainty in the input-output representation of the model arises from uncertainty in the parameter estimates. However, uncertainties in the modeling procedure stem not only from uncertainties in the parameter estimates, but also from measurement errors, from in...

متن کامل

Conjunctive surface–subsurface modeling of overland flow

In this paper, details of a conjunctive surface–subsurface numerical model for the simulation of overland flow are presented. In this model, the complete onedimensional Saint-Venant equations for the surface flow are solved by a simple, explicit, essentially non-oscillating (ENO) scheme. The two-dimensional Richards equation in the mixed form for the subsurface flow is solved using an efficient...

متن کامل

Modeling of Water Flow and Nitrate Transport to Subsurface Drains

In this study, the water flow and nitrate transport to a subsurface drain, using a simplified and detailed model, are simulated for the specific hydro-geological conditions of Elverdinge and Assenede, Belgium. Previously, the DRAIN MOD-N model proved to be able to simulate nitrate concentrations and drainage well for an in-situ leaching experiment, the “Hooibeekhoeve” in the community of Geel (...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: Journal Of Geophysical Research: Solid Earth

سال: 2021

ISSN: ['2169-9356', '2169-9313']

DOI: https://doi.org/10.1029/2020jb020549