Most of the existing correlations for estimating gas viscosity were developed in mid 60's and 70's of the last century. Limited number of data was used to develop them and their accuracies are questionable. Predicting accurate gas viscosity is extremely important in the oil and gas industry as it has a major impact on reservoir recovery, fluid flow, deliverability, and well storage.

In this study, a new correlation has been introduced. This correlation is simpler, features higher accuracy, and uses fewer coefficients compared with the existing correlations. Its application covers a wider range of gas specific gravity without jeopardizing the accuracy of the correlation. Another model was built using Artificial Neural Networks, ANN in order to compare its results with those obtained from the new correlation.

The existing correlations were studied and analyzed using the same, large set of measured data used for this study. Most of these correlations suffered from high errors and thus were optimized using the linear and non-linear regressions. New set of coefficients for these correlations are recalculated for which the accuracy has significantly improved. In spite of such an improvement, the new correlation and new ANN model outperform the existing correlations.

Keywords:
fluid modeling, neural network, artificial intelligence, equation of state, machine learning, upstream oil & gas, spe 153239, dataset 2, viscosity correlation, gas viscosity
Subjects:
Fluid Characterization, Information Management and Systems, Phase behavior and PVT measurements, Fluid modeling, equations of state, Neural networks
Copyright 2012, Society of Petroleum Engineers
You can access this article if you purchase or spend a download.