Simulation of kinetic behavior of natural surfactants adsorption using a new robust approach

Surfactant-based enhanced oil recovery techniques are known as promising methods for mobilizing the trapped oil in porous media. Surfactants can improve oil recovery by modifying the wettability of rock minerals and also by reducing the interfacial tension between injected water and the trapped oil. Natural surfactants have been introduced as good candidates for enhanced oil recovery applications. In addition, they are less expensive and also have less detrimental environmental effects in comparison with the industrial surfactants. Various empirical models have been proposed for simulating the kinetic behavior of surfactants adsorption, but these models suffer from overestimation and underestimation and they cannot be generalized for even a type of surfactant. Therefore, it is crucial to develop a new model that can overcome these issues. In this study, a new simple, rapid, and accurate model based on least square support vector machines (LSSVMs) was developed for predicting the kinetic adsorption density of natural surfactants on both sandstone and carbonate minerals. Coupled simulated annealing algorithm (CSA) is used for tuning the parameters of the model. Predicted values by this model were in an excellent agreement with experimental values with a coefficient of determination of 0.990. The results demonstrated that the proposed LSSVM-CSA model has the best performance in comparison with the other well-established kinetic models. Furthermore, the model reliability was investigated over input parameters changes and showed the acceptable efficiency of the proposed model. A LSSVM-CSA model is proposed for predicting kinetic behavior of natural surfactants adsorption. The proposed model is applicable for both carbonate and sandstone rocks. The model performance is evaluated through statistical and graphical analysis. Results of the LSSVM-CSA model are compared with well-established kinetic models. Results demonstrated that the proposed model exhibits higher accuracy.