Integrating support vector regression with genetic algorithm for CO2-oil minimum miscibility pressure (MMP) in pure and impure CO2 streams

Accurate knowledge of the minimum miscibility pressure (MMP) is essential in successful design of any miscible gas injection process, particularly in CO2 flooding. It is however well-acknowledged that experimental measurements are expensive, time-consuming, and cumbersome. As a direct consequence, a support vector regression model combined with genetic algorithm (GA-SVR) was proposed to predict pure and impure CO2-crude oil MMP. The accuracy and reliability of the proposed model were evaluated through 150 data sets collected in the open literature and compared with approaches commonly used to estimate the MMP (Lee correlation, Shokir correlation, Orr-Jensen correlation, Yellig-Metcalfe correlation, Alston correlation, Emera-Sarma correlation, Cronquist correlation, Kamari et al. correlation, and Fathinasab-Ayatollahi correlation). The results showed that the proposed model for predicting the MMP is in excellent agreement with experimental data and outperforms all the existing methods considered in this work in prediction of pure and impure CO2-oil MMP. Furthermore, outlier diagnosis was performed on the whole data sets to identify the applicable range of all models investigated in this work by detecting the probable doubtful MMP data.