Prediction of supercritical CO₂/brine relative permeability in sedimentary basins during carbon dioxide sequestration

This study aims to accurately determine supercritical CO 2 /brine relative permeability, using a hybrid Genetic Algorithm‐Radial Basis Function (GA‐RBF) neural network. CO 2 sequestration, along with some enhanced oil recovery (EOR) processes, demands an exact knowledge of relative permeability in order to ensure the viability of the operation. Previous studies have shown that errors in CO 2 /brine relative permeability data might result in a four‐fold error in injectivity estimation. This, as well as several recent studies regarding the relative permeability of CO 2 /brine systems, has indicated the importance of this parameter. The developed GA‐RBF model was determined to be in excellent accordance with experimental data, yielding average absolute relative deviations (AARD) of 4.66% and 2.11% for CO 2 and brine relative permeability, respectively. In addition, comprehensive comparisons between classic models and the proposed GA‐RBF model have been carried out. Based on these comparisons, it may be concluded that the proposed model is superior to the classic method (simple correlation) in terms of its accuracy in determining the viability of CO 2 sequestration operations.