Deep-learning-based surrogate flow modeling and geological parameterization for data assimilation in 3D subsurface flow

نویسندگان

چکیده

Data assimilation in subsurface flow systems is challenging due to the large number of simulations often required, and by need preserve geological realism calibrated (posterior) models. In this work we present a deep-learning-based surrogate model for two-phase 3D formations. This model, recurrent residual U-Net (referred as R-U-Net), consists convolutional (convLSTM) neural networks, designed capture spatial-temporal information associated with dynamic systems. A CNN-PCA procedure (convolutional network post-processing principal component analysis) parameterizing complex geomodels also described. approach represents simplified version recently developed supervised-learning-based framework. The R-U-Net trained on simulated saturation pressure fields set random `channelized' (generated using CNN-PCA). Detailed predictions demonstrate that provides accurate results states well responses new realizations, along statistics an ensemble geomodels. procedures are then used combination data problem involving channelized system. Two different algorithms, namely rejection sampling ensemble-based method, successfully applied. overall methodology described paper may enable assessment refinement range realistic problems.

برای دانلود رایگان 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

Reduced-order flow modeling and geological parameterization for ensemble-based data assimilation

Reduced-order modeling represents an attractive approach for accelerating computationally expensive reservoir simulation applications. In this paper, we introduce and apply such a methodology for data assimilation problems. The technique applied to provide flow simulation results, trajectory piecewise linearization (TPWL), has been used previously for production optimization problems, where it ...

متن کامل

Surface-subsurface flow modeling with path-based runoff routing, boundary condition-based coupling, and assimilation of multisource observation data

[1] A distributed physically based model incorporating novel approaches for the representation of surface-subsurface processes and interactions is presented. A path-based description of surface flow across the drainage basin is used, with several options for identifying flow directions, for separating channel cells from hillslope cells, and for representing stream channel hydraulic geometry. La...

متن کامل

Modeling Groundwater Flow in Geological Complex Areas Using Local Grid Refinement, Combined Parameterization

s and proceedings: Vilhelmsen, T.N., Christensen, S., Testing MODFLOW-LGR for simulating flow around buried Quaternary valleys – synthetic test cases, AGU fall meeting 2009, San Francisco, U.S.A., 14-18 December 2009 Vilhelmsen, T.N., Christensen, S., Test af to versioner af MODFLOW-LGR til simulering af grundvandsstrømninger i områder med begravede dale – syntetisk studie, ATV vintermøde 2010,...

متن کامل

Modeling Critical Flow through Choke for a Gas-condensate Reservoir Based on Drill Stem Test Data

Gas-condensate reservoirs contain hydrocarbon fluids with characteristics between oil and gas reservoirs and a high gas-liquid ratio. Due to the large gas-liquid ratio, wellhead choke calculations using the empirical equations such as Gilbert may contain considerable error. In this study, using drill stem test (DST) data of a gas-condensate reservoir, coefficients of Gilbert equation was modifi...

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: Computer Methods in Applied Mechanics and Engineering

سال: 2021

ISSN: ['0045-7825', '1879-2138']

DOI: https://doi.org/10.1016/j.cma.2020.113636