Prediction of Wax Disappearance Temperature by Intelligent Models

It is well-known that reservoir hydrocarbon fluids contain heavy paraffins that may form solid phases of wax at low temperatures. Problems associated with wax formation and deposition are a major concern in production and transportation of hydrocarbon fluids. Thus, testing of wax disappearance temperature (WDT) is essential in high-efficiency development of crude oil. For the sake of reduction of time and improvement of accuracy, four metaheuristic models called gray wolf optimizer-based support vector machine (GWO-SVM), least-squares support vector machine, genetic algorithm-based adaptive network-based fuzzy inference system, and particle swarm optimization-based adaptive network-based fuzzy inference system were used for the prediction of WDT in binary, ternary, and multicomponent systems in the range of 0.1-100 MPa. The input parameters are molar mass and pressure, and the output is the WDT at every point. The comparison between the four models shows that the GWO-SVM gets the best accordance with experimental data sets with the minimum average absolute relative deviation (AARD = 0.7128%), maximum determination coefficient (R-2 = 0.9546), and minimum root-mean-squared error (RMSE = 2.4208) in all 272 data points. And outliers detection using the leverage approach to detect the doubt points, where only 6 data points in all 272 data points.