A Fast and accurate investigation into CO2 Storage challenges by Making a Proxy Model on a Developed Static Model with The Application of Artificial Intelligence/Machine Learning

CO2 emissions as the root of global warming have been intended to be cut by net zero until 2050. CCS is a technology capable of capturing produced CO2 in energy sectors and industries to be injected and stored into the subsurface geological formations like depleted oil and gas reservoirs and aquifers possessing effective trapping mechanisms rather than emissions in the atmosphere. CO2 storage involves drilling an injection well, injection, well control, and CO2 propagation within geological storage, governed by petroleum engineering principles. Therefore, oil and gas companies, besides petroleum engineers, are responsible for the exploration and assessment of viable storages in addition to execution. Among multiple risks, fracturing in caprock and around the wellbore, in addition to leakage through geological paths and legacy wells, are the predominant ones that follow CO2 injection and storage. Storage simulation is employed to monitor well performance and CO2 plume migration to optimize the progress with respect to the objectives and constraints. The common approach is a numerical simulation by commercial and opensource simulators. A full field simulation may take a few hours, depending on the model's type, dimension, and resolution. Sensitivity analysis and optimization of CO2 storage by conventional numerical methods require several runs, which is not temporally efficient. During the last two decades, various types of proxy models have been developed to replicate the reservoir simulation results in the field, well, and grid scales from input features by mathematical, statistical, and AI approaches. The proxy models are able to be substituted for numerical simulators to apply sensitivity analysis, optimization, and history matching extremely quickly without sacrificing accuracy. The most current proxy models are AI and machine learning-based proxy models, called Smart Proxy Model or Surrogate Reservoir Proxy Model (SRM). In this study, after updating the numerical model and designing a feasible injection well, grid-based SRMs were developed to simulate dynamic CO2 saturation and pressure distribution in the entire model's grid blocks in 1 minute with 99% accuracy. Subsequently, the Genetic Algorithm optimized the CO2 injection into the Smeaheia saline aquifer, which is placed in the North Sea, and found the optimum CO2 injection rate and duration, maximizing storage capacity without leakage and fracturing.