A novel model for the solution gas-oil ratio suitable for CO2-rich reservoir fluids

Reservoir fluids with high carbon dioxide (CO2) fraction in their compositions have been the focus of growing attention due to the increasing importance of CO2 injection in subsurface reservoirs for enhanced oil recovery and carbon capture and sequestration. Their thermophysical behavior has unique characteristics when compared to the behavior of reservoir fluids with low CO2 content. However, empirical models that are extensively used to model the reservoir fluid behavior originally did not cover this particular composition, and their extrapolation to this subset of fluids has been little explored in the literature. This study first evaluates the performance of nine such traditional empirical models for the solution gas-oil ratio (Rs) property of CO2-rich reservoir fluids. Predictions are compared against an experimental database containing 1457 points. The database covers reservoir fluids with CO2 molar content ranging from 0% to 38% and gas-oil ratio (GOR) from 75 scf/STB to 2487 scf/STB. Previous models' results show rather poor predictions for reservoir fluids with high CO2 content, with high deviations from experimental data and inconsistencies. Therefore, this paper proposes a general Rs model that is able to model reservoir fluids with high CO2 content and high GOR with significant superior performance. The new Rs model predictions present a mean absolute percentage error (MAPE) of 6% and a root mean square error (RMSE) of 27 scf/STB for the aforementioned experimental values. For a test (holdout) dataset containing 173 data points from novel reservoir fluid compositions, the MAPE was 8% and the RMSE was 71 scf/STB. The proposed model outperforms the existing models on both training and test sets, even for reservoir fluids with low or no CO2 content.