A machine learning approach to subsurface characterization in CO₂-EOR operations

Sequestration of carbon dioxide (CO₂) in deep, geologic formations is one of the most promising solutions to decrease CO₂ concentration in the atmosphere. Depleted hydrocarbon reservoirs are one of the geologic targets considered for sequestration. Even though CO₂ is currently being used in the Enhanced Oil Recovery (EOR) operations, characterization of hydrocarbon reservoirs through an integrated study that includes experiments and numerical simulations is required to evaluate the actual efficiency of the CO₂-EOR operations. A novel Machine Learning algorithm is developed to dynamically characterize hydrocarbon reservoirs. The availability of hydrocarbon reservoirs properties contributes to the optimization of the project, both technically and economically. The results show a good agreement with other experimental literature data.