Abstract:Abstract Geologic CO2 Sequestration (GCS) is a promising engineering technology to reduce global greenhouse emissions. Real-time forecasting of CO2 leakage rates is an essential aspect of large-scale GCS deployment. This work introduces a data-driven, physics-featuring surrogate model based on deep-learning technique for CO2 leakage rate forecasting. The workflow for the development of data-driven, physics-featuring surrogate model includes three steps: 1) Datasets Generation: We first identify uncertainty parameters that affect the objective of interests (i.e., CO2 leakage rates). For the identified uncertainty parameters, various realizations are then generated based on Latin Hypercube Sampling (LHS). High-fidelity simulations based on a two-phase black-oil solver within MRST are performed to generate the objective functions. Datasets including inputs (i.e., the uncertainty parameters) and outputs (CO2 leakage rates) are collected. 2) Surrogate Development: In this step, a time-series surrogate model using long short-term memory (LSTM) is constructed to map the nonlinear relationship between these uncertainty parameters as inputs and CO2 leakage rates as outputs. We perform Bayesian optimization to automate the tuning of hyperparameters and network architecture instead of the traditional trial-error tuning process. 3) Uncertainty Analysis: This step aims to perform Monte Carlo (MC) simulations using the successfully trained surrogate model to explore uncertainty propagation. The sampled realizations are collected in the form of distributions from which the probabilistic forecast of percentiles, P10, P50, and P50, are evaluated. We propose a data-driven, physics-featuring surrogate model based on LSTM for CO2 leakage rate forecasting. We demonstrate its performance in terms of accuracy and efficiency by comparing it with ground-truth solutions. The proposed deep-learning workflow shows promising potential and could be readily implemented in commercial-scale GCS for real-time monitoring applications.

A Science Gateway for the Repeatable Analysis of Machine Learning Predicted Gravity Anomalies

Anomaly Detection in GPS Data Based on Visual Analytics

Gravity Spy: lessons learned and a path forward

Gravity Predictions in Data-Missing Areas Using Machine Learning Methods

Machine Learning-Enhanced Interpolation of Gravity-Assisted Magnetic Data

Inversion of Gravity Anomaly Based on AlexNet Deep Neural Network

Improving access to geodetic imaging crustal deformation data using GeoGateway

Accelerated, Scalable and Reproducible AI-driven Gravitational Wave Detection

Gravity Spy: Integrating Advanced LIGO Detector Characterization, Machine Learning, and Citizen Science

Machine learning for the prediction and Analysis of Bouguer Gravity anomaly of the Kiri uplift in Congo sedimentary Basin

Deep learning for geological hazards analysis: Data, models, applications, and opportunities

A Physics-guided Generative AI Toolkit for Geophysical Monitoring

Inversion of Time-Lapse Surface Gravity Data for Detection of 3D CO$_2$ Plumes via Deep Learning

Algorithms that Defy the Gravity of Learning Curve

Machine Learning Method to Determine Salt Structures from Gravity Data

A Deep Learning Approach to Anomaly Detection in Geological Carbon Sequestration Sites Using Pressure Measurements

Self-Supervised Deep Learning for 3D Gravity Inversion

Deep learning‐based inversion with discrete cosine transform discretization for two‐dimensional basement relief imaging of sedimentary basins from observed gravity anomalies

Improved Gravity Inversion Method Based on Deep Learning with Physical Constraint and Its Application to the Airborne Gravity Data in East Antarctica

CO2 Leakage Rate Forecasting Using Optimized Deep Learning