Optimal Transformations for Multiple Regression: Application to Oil Viscosity Correlation Above and Below Bubble Point Pressures

Abstract

TX 75083-3836, U.S.A., fax 01-972-952-9435. AbstractReservoir oil viscosity is one of the important reservoir fluid properties used for many petroleum engineering calculations such as evaluating hydrocarbon reserves, planning enhanced oil recovery methods, calculating fluid flow rate through reservoir rocks, etc. Therefore, achieving an accurate reservoir oil viscosity value is very crucial for petroleum engineers. Viscosity of reservoir oil can be obtained by a laboratory measurements and empirical correlations. Many of the common correlations in use were developed using data collected from different regions of the world.The literature contains many different viscosity models, the viscosity models found in the literature range from highly theoretical to simple empirical correlations. Two typical viscosity correlations are applicable to both liquid and gas phases; the Corresponding States Method (CSM), semi-empirical, and the empirical Lohrenz-Bray-Clark (LBC) method. In this paper, the applicability of both models were tested to predict the experimental viscosity measurements for pure n-eicosane, ethane/neicosane, propane/n-eicosane binary systems. After normal quality checks, nonlinear multiple regression with linear partial correlation coefficient techniques were applied to develop friction theory and binary interaction parameter were used to modify the representation of friction theory model. The comparisons among LBC, Corresponding State Principle (CSP), general model of friction theory and modified friction theory (using binary interaction parameter) were performed. Based on the error statistical analysis results, the developed friction theory model has shown to provide accurate estimates of viscosity in all pressure regimes.