New models for estimating minimum miscibility pressure of pure and impure carbon dioxide using artificial intelligence techniques

CO2 miscible flooding stands out as one of the most-commonly employed and promising processes in enhanced oil recovery. The key factor for ensuring success in designing and executing CO2 injection projects lies in the precise determination of the Minimum Miscibility Pressure (MMP). Despite significant research endeavors in measuring and predicting the MMP through the use of empirical, analytical, numerical, and experimental approaches, the need for a comprehensive, accurate, efficient, and easy-to-use model remains. In response to this gap, our study developed various predictive models with 205 datasets, including well-established ones such as Artificial Neural Network (ANN), Random Forest (RF), and K-Nearest Neighbors (KNN) along with employing the Gradient Boosting Algorithm (GBA) with the Grid Search (GS) as a novel approach to predict the MMP of both pure and impure CO2. The performance of these models was validated by comparing them with previously published models, using relevant performance metrics such as R2 (coefficient of determination) and MAPE (mean absolute percentage error). Our results demonstrated that the GBA model with GS optimization exhibited remarkable superiority in terms of performance and accuracy in both scenarios, outperforming previously published models as well as the developed ANN, RF, and KNN models. For pure CO2, it achieved an impressive R2 value of 0.999 on the training set and a relatively high value of 0.910 on the validation set. Additionally, the model showed remarkably low MAPE values of 0.0705 % on the training set and 9.238 % on the validation set. Similarly, for impure CO2, the GBA model with GS optimization achieved an R2 value of 0.999 on the training set and a relatively high value of 0.865 on the validation set. The model also exhibited low MAPE values of 1.66 % on the training set and 6.798 % on the validation set. These findings highlight the potential of the GBA model with GS optimization as a valuable approach for accurately estimating the MMP, offering a cost-effective and efficient approach for designing and optimizing CO2 flooding operations in enhanced oil recovery processes.