Perpustakaan Digital ITB

Surfactant injection is gaining attention for producing oil from reservoirs. Producing oil from reservoirs with high wax content is challenging because it usually exhibits high interfacial tension (IFT) with brine reservoirs. Surfactants can reduce IFT by adsorbing at the oil-water interface, increasing waxy-oil recovery. However, selecting high-performance surfactants is a time-consuming and high-cost process. Therefore, this study aims to comprehensively analyze and optimize the surfactant screening process by utilizing a machine-learning model to predict the optimum salinity that resulted in low interfacial tension, including the effect of hydrophobic number, hydrophilic group, type of surfactant, EACN of oil, concentration, temperature, and oil solubilization ratio. Salinity can be used as a property for surfactant selection because it affects IFT, where the increase in salinity indicates a decrease in IFT. The data were limited laboratory test datasets (phase behavior and IFT value) from three single and three surfactant mixtures. The predictive machine learning model was developed using several approaches, such as Gradient-Boosted Trees, Random Forests, and Tree Ensembles, to estimate optimum surfactant salinity, which results in low IFT. The best predictive model obtained is the Gradient-Boosted Tree with model performance (R2=0.911, MAE=0.19, RMSE=0.284) confirming the acceptable accuracy of the developed correlation for salinity prediction. However, further studies are needed to test this model using various databases of other surfactants to validate that this model is reliable.