Evaluation of different artificial intelligent models to predict reservoir formation water density

Nearly all hydrocarbon reservoirs are bounded by water-saturated rocks, namely aquifers. In addition to natural water drive, there is an artificial water drive mechanism in which water is injected into formation to intensify the reservoir pressure. This method, employed to induce the hydrocarbon production, is called water flooding. Several laboratory researches have shown that oil recovery can be heightened by making some alterations to injected brine salinity through water flooding. Accordingly, acquiring exact information about the PVT characteristics of brine is necessary. Density is a property of great importance as it is employed in various physical, chemical, geothermal, and geochemical aspects. The authors aimed to develop a dependable intelligent method to accurately predict the brine density at elevated temperatures and pressures. MLP and GA-RBF models were utilized in this study. The results showed that the proposed model is capable of accurately predicting the brine density at elevated pressures and temperatures for different concentrations of brine. The correlation factor of 1.0000 and root mean squared error of 3.27E-05 demonstrate the accuracy of the proposed model.