Mathematical Modelling of Oil Viscosity at Bubble point Pressure and Dead Oil Viscosity of Nigerian Crude

Abstract

Traditionally, reservoir engineering fluid flow calculations use viscosity data. However, in the absence of lab/experimental data other available derived correlations are used to predict the PVT property. Limit on the number of available data, regional peculiarity of the fluid, several viscosity correlations in the literature have limited accuracy and applicability. This study has developed predictive models using more than 2020 unpublished PVT data sets from different locations in Nigeria in rigorous nonlinear regression modelling. Different nonlinear algorithms, modified Newton-Raphson nonlinear least-square data fitting approach; Levenberg-Marquardt algorithm were used to develop new models for the estimation of the viscosity at the bubblepoint pressure and dead oil viscosity. The results of the performance of the model for viscosity at the bubblepoint show that the model provides better prediction with average absolute relative error of 21.06 and coefficient of correlation of 0.98 and the dead oil viscosity model shows a substantial improvement with average absolute relative error of 30.06 and coefficient of correlation of 0.90 over published correlations.