Skip to content

A Comparison of Methods for Creating Flow Zone Units in the Gippsland Basin, Australia

Filipenko, David

Engineering Honours 2007

University of Adelaide

Abstract

Modelling fluid flow through reservoir rock has traditionally involved a linear plot between log permeability versus porosity; this relationship does not account for data with multiple permeabilities for a given porosity. The concept of Flow Zone Units (FZU) addresses the presence of multiple permeabilities for a given porosity, and provides a mathematically rigorous method based on the Carman–Kozeny equation for characterising rocks based on their flow properties.

The study was conducted over the Halibut and Fortescue Fields in the eastern Gippsland Basin, Australia, with data being in the form of well completion reports (WCR) and digital log data. These data sets were then analysed using three FZU methods, graphical clustering, analytical clustering, and the Biniwale–Behrenbruch method which integrates geological facies interpretation into the FZU study.

Resulting from the study, the Biniwale–Behrenbruch method was found to be the most insightful when attempting to map FZUs within cores, as it required the integration of a large amount of data from various sources. Sands from both Halibut-1 and Halibut-2 were successfully analysed using the Biniwale–Behrenbruch method. However, an absence of detailed information in the WCR made a complete Biniwale–Behrenbruch analysis impossible.

The Analytical Method, using Ward’s Algorithm, was the most successful of the mechanistic methods (i.e. Graphical Clustering and the Analytical Method) due to the way that Ward’s Algorithm clusters data. While significantly less data integration is needed when comparing to the Biniwale–Behrenbruch method, Flow Zone Types can easily be determined for wells with fewer data points making this a method suitable for the Fortescue and Halibut fields.

A FZT prediction using well logs was performed using the Analytical Approach. Flow Zone Type prediction using well logs was made difficult as depth correlation between the well logs and core data was not performed. The methods trialled were iterative multivariate regression and artificial neural networks; however errors in the results were often larger than the variance in the results making any prediction meaningless.


Australian School of Petroleum
THE UNIVERSITY OF ADELAIDE

SA 5005 AUSTRALIA

Contact

T: +61 8 8313 8000
F: +61 8 8313 8030
email