2015-Sustainable Industrial Processing Summit
SIPS 2015 Volume 7: Ionic Liquids & Energy Production
| Editors: | Kongoli F, Gaune-Escard M, Mauntz M, Rubinstein J, Dodds H.L. |
| Publisher: | Flogen Star OUTREACH |
| Publication Year: | 2015 |
| Pages: | 310 pages |
| ISBN: | 978-1-987820-30-0 |
| ISSN: | 2291-1227 (Metals and Materials Processing in a Clean Environment Series) |
< CD shopping page
Application of Artificial Neural Networks Technique for Estimating Porosity and Permeability in an Oil Field
Abdollah Esmaeili1;1NATIONAL IRANIAN OIL COMPANY (N.I.O.C), Omidieh, Iran (Islamic Republic of Iran);
Type of Paper: Regular
Id Paper: 104
Topic: 17
Abstract:
Porosity and permeability are the most important hydrocarbon reservoir parameters. There are two methods for porosity determination including direct (core analysis by helium injection)and indirect (well log analysis). Similarly, permeability can be obtained at the laboratory from core samples by dry air injection or well testing method. These methods demand high cost and are time consuming. So, due to economic reasons and impossibility of coring in horizontal wells, core data would be available for limited number of wells. However, most wells have well log data. In the present study, intelligent soft computing neural networks that nowadays are widely used in petroleum industry were used for prediction of porosity and permeability in an oil field. In this study, MATLAB software was used to handle neural networks core and well logs data, including porosity and permeability. These networks were developed using error-back propagation algorithm within feed-forward networks. After comparing the measured and network predicted results, the parameters of the Artificial Neural Network (ANN) were adjusted for a desired network. The correlation coefficient obtained between core results and ANN predicted porosity and permeability are 0.82 and 0.92, respectively. These results show that intelligent neural network models have predicted porosity and permeability successfully. Finally, the above mentioned networks were generalized to the third well which had no core data.