A physics-constraint neural network for CO2 storage in deep saline aquifers during injection and post-injection periods
Mengjie Zhao, Yuhang Wang, Marc Gerritsma, Hadi Hajibeygi
Abstract
CO 2 capture and storage is a viable solution in the effort to mitigate global climate change. Deep saline aquifers, in particular, have emerged as promising storage options, owing to their vast capacity and widespread distribution. However, the task of proficiently monitoring and simulating CO 2 behavior within these formations poses significant challenges. To address this, we introduce the physics-constraint neural network for CO 2 storage (CO 2 PCNet), a model specifically designed for simulating and monitoring CO 2 storage in deep saline aquifers during injection and post-injection periods. Recognizing the significant challenges in accurately modeling the distribution and movement of CO 2 under varying permeability conditions, the CO 2 PCNet integrates the principles of physics with the robustness of deep learning, serving as a powerful surrogate model. The architecture of CO 2 PCNet starts with an encoder that adeptly processes spatial features from overall mole fraction ( z CO 2 ) and pressure fields ( P l ), capturing the complex dynamics of a CO 2 trajectory. By incorporating permeability information through a conditioning step, the network ensures a faithful representation of the influences on CO 2 behavior in subsurface conditions. A ConvLSTM module subsequently discerns temporal evolutions, reflecting the real-world progression of CO 2 plumes within the reservoir. Lastly, the decoder precisely reconstructs the predictive spatial profile of CO 2 distribution. CO 2 PCNet, with its integration of convolutional layers, recurrent mechanisms, and physics-informed constraints, offers a refined approach to CO 2 storage simulation. This model offers the potential of utilizing advanced computational methods in advancing CCS practices. • A novel physics-constraint neural network (CO 2 PCNet) is proposed for dynamic CO 2 storage. • ConvLSTM integration captures dynamic spatial–temporal changes during full-cycle process. • Incorporating PDE constraints ensures predictions match physical laws. • Comparisons of CO 2 PCNet with baselines demonstrate accurate, long-term CO 2 storage predictions.