Further model development for prediction of reservoir oil viscosity

Viscosity of oil is a crucial parameter in various aspects of oil and gas industry including reservoir simulation, production analysis and equipment design problems. Therefore, proposing simple, robust and universal approaches for estimation of oil viscosity is very important. In the current study a...

Full description

Saved in:
Bibliographic Details
Published in:Petroleum science and technology 2022, Vol.40 (3), p.310-321
Main Authors: Mokarami Moghadam, Elham, Naseri, Ali, Riahi, Mohammad Ali
Format: Article
Language:English
Subjects:
Model accuracy
multilayer perceptron
Oil fields
prediction
PVT
reservoir fluid
Reservoirs
Viscosity
Viscosity measurement
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Viscosity of oil is a crucial parameter in various aspects of oil and gas industry including reservoir simulation, production analysis and equipment design problems. Therefore, proposing simple, robust and universal approaches for estimation of oil viscosity is very important. In the current study a computer-based model namely MLP-NN was designed for estimation of viscosity of dead oil, saturated oil and under saturated oil. A number of 1900 laboratory measured viscosity data of Iranian oilfields associated with wide variations of operational variables were considered to construct the model. The performance accuracy of the method was assessed by considering a number of performance measure statistical variables and results comparison with other available predictive methods. The AAPRE% values of the model for viscosity of dead oil, saturated oil, and under saturated oil were 13.02, 2.45, and 10.55, respectively. The model also provided R 2 values of 0.989, 0.9953, and 0.9883 for viscosity of dead oil, saturated oil, and under saturated oil. The outcomes indicated the precision and reliability of the implemented method and its superiority over available predictive methods.
ISSN:1091-6466
1532-2459
DOI:10.1080/10916466.2021.1993914