|
چکیده
|
Accurate determination of interfacial tension (IFT) between crude oil and brines is crucial in enhanced oil recovery (EOR) processes. However, the available IFT models are only applicable to systems containing pure hydrocarbons and saline waters. The current study aims at designing comprehensive predictive tools for the IFT between asphaltenic crude oils and various brines. Hence, 339 relevant experimental data, covering an extensive range of operating conditions, were gathered from literature. After determining the most effective input variables through the Spearman's rank coefficient, the experimental data were employed to train the smart soft-computing approaches, i.e., radial basis function (RBF), multilayer perceptron (MLP) and gaussian process regression (GPR). Although all novel predictive tools presented excellent results, the one designed based on the GPR method was recognized as the most reliable model with the average absolute relative error (AARE) of 0.67% and R2 value of 99.63%, in the testing (validation) stage. Additionally, it estimated the majority of the IFT data with relative errors below 0.10%. On the other hand, the validity of gathered databank was confirmed through the leverage method. The influences of pressure, temperature, salinity and structural characteristics of salts on the IFT were discussed in details, and the proposed models favorably described the physical trends. Eventually, a sensitivity analysis was carried out based on the GPR model to clarify the order of significance of factors in controlling the crude oil – brine IFT.
|