AshEse Journal of Engineering
Vol. 1(1), pp. 001-007, April, 2015
© 2015 AshEse Visionary Limited
Full Length Research
Sadegh Baziar1*andHabibollah Bavarsad Shahripour2
1Department of Petroleum Engineering, Amirkabir University of Technology, Tehran, Iran
2International Petro Offshore Niam Kish, Tehran, Iran
Received March 2015; Accepted April 2015
This paper presents a new correlation approach to predict total formation volume factor below the bubble point pressure for oil and gas mixtures. This correlation is obtained by using more than 450 experimental data points which are collected from samples of Iranian oil reservoirs. The important factors of influencing parameters are determined using an artificial neural network. Then an appropriate form of correlation is developed by multivariable regression. Finally by use of nonlinear optimization, the correlation coefficients are adjusted in an optimum level to minimize average absolute relative error. This new correlation is valid in a broad range of pressure and temperature and is more accurate than other ones for Iranian oil mixtures.
Key words:Artificial neural network, correlation, multivariable regression, nonlinear optimization, total formation volume factor.