Application of constrained multi-variable search methods for prediction of PVT properties of crude oil systems

• Published in: Fluid phase equilibria 2014-02, Vol.363, p.121-130
• Main Authors: Arabloo, Milad, Amooie, Mohammad-Amin, Hemmati-Sarapardeh, Abdolhossein, Ghazanfari, Mohammad-Hossein, Mohammadi, Amir H.
• Format: Article
• Language: English
• Subjects: Bubble point pressure; Group analysis; New models; Oil formation volume factor; Statistical quality measures
• ISSN: 0378-3812; 1879-0224
• DOI: 10.1016/j.fluid.2013.11.012

•Improved models for prediction of PVT properties of reservoir oil systems have been developed.•More than 750 data series were gathered from different geographical locations worldwide.•SLP was incorporated in addition to GRG algorithms to expedite the process of achieving a good feasible solution.•Comparative studies have been carried out between developed models and other published correlations.•The developed models are accurate, reliable and superior to all other published correlations. Accurate prediction of the PVT properties of reservoir oil is of primary importance for improved oilfield development strategies. Experimental determination of these properties is expensive and time-consuming. Therefore, new empirical models for universal reservoir oils have been developed as a function of commonly available field data. In this communication, more than 750 experimental data series were gathered from different geographical locations worldwide. Successive linear programming and generalized reduced gradient algorithm as two constrained multivariable search methods were incorporated for modeling and expediting the process of achieving a good feasible solution. Moreover, branch-and-bound method has been utilized to overcome the problem of stalling to local optimal points. In-depth comparative studies have been carried out between the developed models and other published correlations. Finally, a group error analysis was performed to study the behavior of the proposed models as well as existing correlations at different ranges of independent variables. It is shown that the developed models are accurate, reliable and superior to all other published correlations.