Abstract

Substantial amounts of gas are accumulated in tight sand reservoirs in Cooper basin, Australia. This basin has a complex geomechanical stress regime. It is believed that the orientation of the minimum horizontal stress is changing and also there is a stress dependent permeability in this basin. These characteristics have a major impact on the gas recovery from different fields in Cooper basin.

It is well known that permeability decreases in sandstone reservoirs when the net confining stress on the rock increases as the gas pressure in the reservoir decreases, so permeability is reduced by gas production. This permeability reduction is more pronounced in tight sand reservoirs due to shear dilation of micro-fractures. In order to model stress dependent permeability, a fully coupled finite element geomechanical reservoir simulations is needed. However, since a large number of data and high computational time is required for this kind of modeling, stress dependent permeability model is a short cut for reservoir simulation. This paper presents an alternative approach of partial coupling of geomechanics and reservoir simulation in which only a finite difference simulator is used with much less computational time and lower cost.

In order to prove stress dependent permeability and develop a model, first a three dimensional hydraulic fracture model is constructed based on wire-line logs and mini-frac tests, then the resulting fracture geometry and conductivity are used as input data for dynamic reservoir flow simulation. We found that the reservoir permeability is changing at different time periods during pressure and production history matching.

In this paper a stress-permeability relationship was developed by assuming a fully elastic fractured network of flow path to the wellbore in a hydraulically fractured well. Then, this model was tested by a partially coupled geomechanical simulation in Big Lake field in Cooper basin. The results show that our approach has better predictions than when the permeability is assumed to be constant. By this procedure not only the predictions are improved but also the computational time is reduced.

You can access this article if you purchase or spend a download.