Modeling minimum miscibility pressure during pure and impure CO 2 flooding using hybrid of radial basis function neural network and evolutionary techniques

Abstract Minimum miscibility pressure (MMP) is the most significant parameter monitoring the efficiency of CO 2 flooding by establishing the miscibility condition in the oil reservoirs resulting in a higher ultimate oil recovery factor. To date, considerable investigations on CO 2 -MMP determination have been implemented; however, developing more universal models is still needed. In the present study, a number of network-based strategies, named as radial basis function neural network optimized with five evolutionary algorithms (RBF-EAs); namely genetic algorithm (GA), particle swarm optimization (PSO), imperialist competitive algorithm (ICA), ant colony optimization (ACO), and differential evolution (DE), were developed for estimating pure/impure CO 2 -MMP. The most comprehensive source of data including about 270 CO 2 -MMP values was utilized for RBF modeling. Crossplot, cumulative frequency diagram, and trend analysis as visual tools, and root mean square error (RMSE), average absolute percent relative error (AAPRE) and determination coefficient (R 2 ) as the statistical parameters, were utilized in this study to evaluate the comprehensiveness of the developed RBF tools. It was found that the ICA-RBF model is the most accurate method with statistical values of RMSE = 1.16, R 2  = 0.95 and AAPRE = 6.01%. The ICA-RBF method is more accurate than the best smart methodology developed in the literature with respect to the AAPRE parameter. Besides, temperature can be considered as the most affecting input data on the MMP estimations because of sensitivity analysis implemented here. In summary, the ICA-RBF mathematical strategy can provide a rapid and reasonably accurate prediction of MMP during the injection of both pure and impure streams of CO 2 . The proposed strategy in this study is of paramount weight for engineers and scientist working on enhanced oil recovery.