Abstract:Geological carbon storage entails the injection of megatonnes of supercritical CO2 into subsurface formations. The properties of these formations are usually highly uncertain, which makes design and optimization of large-scale storage operations challenging. In this paper we introduce a history matching strategy that enables the calibration of formation properties based on early-time observations. Early-time assessments are essential to assure the operation is performing as planned. Our framework involves two fit-for-purpose deep learning surrogate models that provide predictions for in-situ monitoring well data and interpreted time-lapse (4D) seismic saturation data. These two types of data are at very different scales of resolution, so it is appropriate to construct separate, specialized deep learning networks for their prediction. This approach results in a workflow that is more straightforward to design and more efficient to train than a single surrogate that provides global high-fidelity predictions. The deep learning models are integrated into a hierarchical Markov chain Monte Carlo (MCMC) history matching procedure. History matching is performed on a synthetic case with and without 4D seismic data, which allows us to quantify the impact of 4D seismic on uncertainty reduction. The use of both data types is shown to provide substantial uncertainty reduction in key geomodel parameters and to enable accurate predictions of CO2 plume dynamics. The overall history matching framework developed in this study represents an efficient way to integrate multiple data types and to assess the impact of each on uncertainty reduction and performance predictions.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to use early - observation data (such as 4D seismic data and monitoring well data) to calibrate formation properties during geological carbon storage, thereby reducing uncertainties and improving the prediction accuracy of CO₂ reservoir dynamics. Specifically, the researchers developed a deep - learning - based history - matching framework, aiming at: 1. **Reducing Uncertainties**: By combining 4D seismic data and monitoring well data, reduce the uncertainties of key geological model parameters, so as to predict the dynamics of CO₂ plumes more accurately. 2. **Efficient Data Assimilation**: Under highly uncertain prior conditions, use deep - learning surrogate models to efficiently perform data assimilation, especially in early - time predictions when uncertainties are high but performance evaluation is crucial. 3. **Multi - scale Data Fusion**: 4D seismic data and monitoring well data have different resolutions. Therefore, the researchers constructed two specialized deep - learning networks to process these data respectively, making the entire workflow more concise and the training more efficient. ### Main Methods and Contributions - **Deep - learning Surrogate Models**: The researchers developed two deep - learning surrogate models, one for predicting the interpreted 4D seismic saturation data, and the other for predicting the vertical saturation profiles of monitoring wells. These two models use 3D U - Net and 1D U - Net architectures respectively. - **Hierarchical MCMC History Matching**: Integrate the deep - learning surrogate models into a hierarchical Markov Chain Monte Carlo (MCMC) history - matching procedure to evaluate the impact of different data types on uncertainty reduction. - **Synthetic Case Analysis**: Through synthetic cases, the researchers demonstrated the effect of history - matching with and without 4D seismic data, and quantified the impact of 4D seismic data on uncertainty reduction. ### Formula Summary - **Geological Model Parameters**: \[ h = [\mu_{\log k}, \sigma_{\log k}, \log_{10} a_r, d, e] \] where: - \(\mu_{\log k}\) and \(\sigma_{\log k}\) represent the mean and standard deviation of the permeability logarithmic field respectively. - \(a_r\) is the anisotropy ratio of vertical and horizontal permeabilities. - \(d\) and \(e\) are parameters related to the relationship between porosity and permeability. - **Relationship between Porosity and Permeability**: \[ (\phi_s)_i = d \cdot (\log k_s)_i + e, \quad i = 1, \ldots, n_s \] - **Loss Functions**: - **Loss Function of Seismic Data Surrogate Model**: \[ L_{\text{seis}}(\theta_{\text{seis}}) = \frac{1}{n_{\text{seis}}^x n_{\text{seis}}^y n_{\text{seis}}^z n_t} \sum_{i = 1}^N \| \hat{S}_{\text{seis},i} - S_{\text{seis},i} \|_2^2 \] - **Loss Function of Monitoring Well Data Surrogate Model**: \[ L_{\text{mon}}(\theta_{\text{mon}}) = \frac{1}{n_z n_m^t} \sum_{i = 1}^N \| \hat{S}_{\text{mon},i} - S_{\text{mon},i} \| \] Through these methods, the researchers provided an efficient and accurate history - matching framework that can effectively use multiple data types to reduce uncertainties in geological carbon storage and improve prediction accuracy.

Deep Learning Framework for History Matching CO2 Storage with 4D Seismic and Monitoring Well Data

History Matching for Geological Carbon Storage using Data-Space Inversion with Spatio-Temporal Data Parameterization

Surrogate Model for Geological CO2 Storage and Its Use in Hierarchical MCMC History Matching

Accelerated training of deep learning surrogate models for surface displacement and flow, with application to MCMC-based history matching of CO2 storage operations

Deep Learning-Based Geological Parameterization for History Matching CO2 Plume Migration in Complex Aquifers

Uncertainty Quantification Based on Deep-Learning Approach Integrating Time-Lapse Seismic Data for Geological Carbon Storage

A deep learning-accelerated data assimilation and forecasting workflow for commercial-scale geologic carbon storage

A Robust Deep Learning Workflow to Predict Multiphase Flow Behavior during Geological CO2 Sequestration Injection and Post-Injection Periods

Integrated Carbonate Lithofacies Modeling Based on the Deep Learning and Seismic Inversion and Its Application

An efficient deep learning-based workflow for CO 2 plume imaging considering model uncertainties with distributed pressure and temperature measurements

Deep learning for characterizing CO 2 migration in time-lapse seismic images

Connect the Dots: In Situ 4D Seismic Monitoring of CO2 Storage with Spatio-temporal CNNs

Deep Learning-Accelerated 3d Carbon Storage Reservoir Pressure Forecasting Based on Data Assimilation Using Surface Displacement from Insar

Deep learning-aided image-oriented history matching of geophysical data

Reinforcement learning based automated history matching for improved hydrocarbon production forecast

Likelihood-Free Inference and Hierarchical Data Assimilation for Geological Carbon Storage

History Matching Complex 3D Systems Using Deep-Learning-Based Surrogate Flow Modeling and CNN-PCA Geological Parameterization

Joint inversion of Time-Lapse Surface Gravity and Seismic Data for Monitoring of 3D CO$_2$ Plumes via Deep Learning

Assimilation of Geophysics-Derived Spatial Data for Model Calibration in Geologic CO2 Sequestration

An Ensemble 4D Seismic History Matching Framework with Sparse Representation Based on Wavelet Multiresolution Analysis