Mapping of seismic and lithologic facies from 3D reflection data plays a key role in depositional environment analysis reservoir characterization during hydrocarbon exploration development. Although variety machine-learning methods have been developed to speed up interpretation improve prediction accuracy, there still exist significant challenges multiclass classification practice. Some these limitations include complex representation, limited training with labels, imbalanced class...

Abstract Geophysical monitoring of geologic carbon sequestration is critical for risk assessment during and after dioxide (CO 2 ) injection. Integration multiple geophysical measurements a promising approach to achieve high‐resolution reservoir monitoring. However, joint inversion large data challenging due high computational costs difficulties in effectively incorporating from different sources with resolutions. This study develops differentiable physics model large‐scale inverse problems...

Among the large variety of mathematical and computational methods for estimating reservoir properties such as facies petrophysical variables from geophysical data, deep machine-learning algorithms have gained significant popularity their ability to obtain accurate solutions inverse problems in which physical models are partially unknown. Solutions classification inversion generally not unique, uncertainty quantification studies required quantify model predictions determine precision results....

Abstract Computation of petrophysical properties on digital rock images is becoming important in geoscience. However, it usually complicated for natural heterogeneous porous media due to the presence multiscale pore structures. To capture heterogeneity rocks, we develop a method based deep generative adversarial networks assimilate imaging data generation synthetic high‐resolution rocks having large field view. The reconstructed not only honor geometric structures 3‐D micro‐CT but also...

We have developed a new stochastic nonlinear inversion method for seismic reservoir characterization studies to jointly estimate elastic and petrophysical properties quantify their uncertainty. Our aims multiple realizations of the entire set properties, including velocities, density, porosity, mineralogy, saturation, by iteratively updating initial ensemble models based on mismatch between response measured data. The are generated using geostatistical methods geophysical forward operators...

Estimating rock and fluid properties in the subsurface from geophysical measurements is a computationally memory-intensive inverse problem. For nonlinear problems with non-Gaussian variables, analytical solutions are generally not available, of those must be approximated using sampling optimization methods. To reduce computational cost, model data can reparameterized into low-dimensional spaces where solution problem computed more efficiently. Among potential dimensionality reduction...

Abstract Effective permeability is a key physical property of porous media that defines its ability to transport fluid. Digital rock physics (DRP) combines modern tomographic imaging techniques with advanced numerical simulations estimate effective properties. DRP used complement or replace expensive and time‐consuming impractical laboratory measurements. However, increase in sample size capture multimodal multiscale microstructures, conventional approaches based on direct simulation (DNS)...

We have developed a time-lapse seismic history matching framework to assimilate production data and for the prediction of static reservoir models. An iterative assimilation method, ensemble smoother with multiple is adopted iteratively update an models until their predicted observations match actual measurements quantify model uncertainty posterior To address computational numerical challenges when applying ensemble-based optimization methods on large volumes, we develop deep representation...

The main challenge in the inversion of seismic data to predict petrophysical properties hydrocarbon-saturated rocks is that physical relations link model often are nonlinear and solution inverse problem generally not unique. As a possible alternative traditional stochastic optimization methods, we develop method adopt machine-learning algorithms by estimating between unknown variables from training set with limited computational cost. We probabilistic approach for based on physics-informed...

Seismic facies analysis interprets depositional environment and types from the reflection seismic data, an important step in exploration reservoir characterization. While machine learning methods, especially deep models such as convolutional neural networks (CNNs) have been applied to assist interpretation salt identification, significant challenges still remain for 3D multi-class classification: complex data representation, limited labeled training, imbalanced class distribution lack of...

Carbon dioxide sequestration in deep saline aquifers and depleted reservoirs relies on numerical models for the prediction of spatial distribution CO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> saturation during injection migration. Due to limited knowledge rock fluid properties before injection, model predictions are often uncertain must be updated when new measurements available. The plume location can monitored using time-lapse...

In this work, we propose an ensemble-based seismic history matching approach to predict reservoir properties, i.e. porosity and permeability, with uncertainty quantification, using both production time lapse data. To avoid the common underestimation of in ensemblebased optimization approaches, make computation feasible, introduce convolutional autoencoder reparameterize data into a lower dimensional space. We then apply Ensemble Smoother Multiple Data Assimilation optimize ensemble models...

Abstract Geophysical subsurface characterization plays a key role in the success of geologic carbon sequestration (GCS). While deterministic inversion methods are commonly used due to their computational efficiency, they often fail adequately quantify model uncertainty, which is essential for informed decision‐making and risk mitigation GCS projects. In this study, we propose SVGD‐AE method, novel geostatistical approach that integrates geophysical data with prior geological knowledge...

Abstract Accurate calculation of adsorbed shale gas content is critical for reserve evaluation and development. However, adsorption desorption experiments are expensive time-consuming, while physics-based models empirical correlations unable to accurately capture the characteristics different shales. Langmuir one most commonly used model calculating in reservoirs. existing pressure volume oversimplified based on limited experimental data points. Thus they not representative key geological...

Determining effective elastic properties of rocks from their pore-scale digital images is a key goal rock physics (DRP). Direct numerical simulation (DNS) behavior, however, incurs high computational cost; and surrogate machine learning (ML) model, particularly convolutional neural network (CNN), show promises to accelerate homogenization process. 3D CNN models, are unable handle large due memory issues. To address this challenge, we propose novel method that combines with hierarchical...

The petrophysical inversion of seismic data is one the key components reservoir characterization. goal to estimate properties rocks, such as porosity, volumes minerals or lithologies, and water hydrocarbon saturations, from data. This process can be performed by combining amplitude-variation-with-offset (AVO) modeling rock-physics relations with deterministic probabilistic inverse theory methods, either in a two-step approach based on AVO mapping single-loop step that includes both. We...

Earth and Space Science Open Archive This work has been accepted for publication in Geophysical Research Letters. Version of RecordESSOAr is a venue early communication or feedback before peer review. Data may be preliminary. Learn more about preprints. preprintOpen AccessYou are viewing the latest version by default [v1]Multiscale fusion digital rock images based on deep generative adversarial networksAuthorsMingliangLiuiDTapanMukerjiSee all authors Mingliang LiuiDCorresponding Author•...

基于2004年10月对全湖67个采样点水下光合有效辐射(photosynthetically active radiation:PAR)和各光学活性物质浓度的测定,分析了真光层深度的空间分布及其影响因素.利用实测的叶绿素a浓度,真光层深度,PAR强度,由水温计算得到的最佳固碳速率以及由经纬度计算的日照周期等,在垂向归纳模型(vertically generalized production...