Generative Adversarial Network for Modeling of CO2 Plume Evolution in Geological Carbon Storage Systems

Abstract Deep learning-based surrogate models offer a powerful alternative to traditional numerical simulations for tackling subsurface multiphase flow challenges, such as those in Geological Carbon Storage (GCS). In this work, we utilized a deep learning surrogate model using Generative Adversarial Networks (GANs) to model CO2 saturation and pressure buildup evolution. GANs, with their generator-discriminator framework, excel at capturing intricate spatial and temporal patterns, enabling accurate predictions of CO2 plume saturation and pressure buildup in saline aquifers. By leveraging adversarial training, GANs effectively model structured data, preserving spatial relationships and generating high-resolution outputs. We first developed physics-based numerical simulation models to represent both the injection and post-injection phases of GCS. Using Latin-Hypercube sampling, we created a diverse set of reservoir and decision parameters, forming a comprehensive simulation database. The GANs were trained and evaluated on a 2D model to assess their performance. During training, we employed Mean Squared Error (MSE) and spatial derivative-based loss functions to optimize the model hyperparameters. The GANs demonstrated strong performance, achieved R2 values of 0.989 and 0.996 for saturation and pressure buildup predictions, respectively. The Normalized Absolute Error (NAE) remained consistently around 1% across all predictions, highlighting the model’s accuracy in capturing the temporal and spatial evolution of CO2 saturation. Additionally, the GANs demonstrated remarkable computational efficiency, with prediction times of just 0.01 seconds per case, compared to 1000 seconds for the 2D model using physics-based simulations. These results underscore the potential of GANs to deliver highly accurate predictions while significantly reducing computational costs.