Ant Colony Optimization Algorithm of Solution Gas Oil Ratio PVT Correlation

Abstract

The solution gas-oil ratio (Rs) is a vital parameter in pressure-volume-temperature (PVT) analysis and reservoir engineering, with its accurate estimation being critical during reservoir depletion. While laboratory-based fluid sampling offers precision, it is often capital-intensive and logistically demanding. Consequently, empirical correlations have been developed to estimate Rs; however, these correlations are frequently region-specific and fail to generalize across diverse reservoir conditions, leading to significant prediction errors. This study aimed to optimize existing Rs correlations—specifically the Glaso (1980) and Standing (1947) models—using the Ant Colony Optimization (ACO) algorithm to enhance prediction accuracy for the Volve field. Data from the Volve production dataset underwent extensive cleaning to remove irrelevant features, missing entries, and anomalous zero values, ensuring reliability for modeling. The ACO algorithm was then applied to calibrate the parameters of the selected correlations, with optimization assessed using statistical metrics such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and the Correlation Coefficient (R). Results showed that the optimized Glaso correlation achieved superior performance, yielding a high correlation coefficient (0.9993) and a significant reduction in average relative error (30.22%), outperforming its original and the Standing model in predictive accuracy. Comparative analysis against experimental data and traditional models confirmed the robustness and adaptability of the ACO-enhanced approach. Despite challenges such as data dependency, parameter sensitivity, and risk of overfitting, the ACO algorithm demonstrated strong potential for improving Rs estimation across complex reservoir systems. The findings underscore the necessity of optimizing empirical models before their field application and affirm the value of bio-inspired algorithms in petroleum reservoir analysis.