Assessment of CO2 storage potential in reservoirs with residual gas using deep learning

CO2 injection into the underlying water leg of depleted hydrocarbon reservoirs is a desirable option for carbon storage as demonstrated by existing industrial-scale storage projects in these geologic environments. This study sheds light on the effect of residual methane on the CO2 storage efficiency as a screening criterion for selecting a water-bearing zone of a depleted gas reservoir to store CO2. Using compositional reservoir simulations, we have evaluated the impact of residual methane on the injectivity, operational pressure, and long-term CO2 trapping efficiency during injection and postinjection stage in a reservoir model representative of the so-called HC sand gas reservoir in the High Island 24L field located in the offshore Texas State Waters. Results suggest that the presence of residual hydrocarbon gas negatively affects CO2 residual and dissolution trapping because it enhances the injectivity and pressure management arising from the increased mobility of CO2 plume in the vicinity of the injection zone due to its mixing with the resident residual hydrocarbon gas. We further investigate the application of artificial neural network (ANN)-based proxy models for fast-track modeling of CO2 storage in geologic structures associated with depleted gas reservoirs, aiming at the prediction of CO2 trapping efficiency. We then use the developed ANN model to perform Monte Carlo simulations for quantifying the uncertainty of geologic and reservoir parameters on CO2 trapping efficiency in these formations. It becomes evident that the residual hydrocarbon saturation is a key screening criterion for the storage site selection. The developed data-driven model can offer a robust and fast tool for screening the water-bearing zone of the depleted gas reservoirs by evaluating the efficiency of CO2 storage.

Automated summary

This study shows that residual methane negatively affects CO2 storage efficiency and proposes an artificial neural network-based model for screening water-bearing zones in depleted gas reservoirs to evaluate CO2 storage efficiency.