"Predictive Model for Permeability Reduction by Small Wetting Phase Saturations"

Field experience in unconventional gas reservoirs indicates that well deliverability can vary dramatically, even between closely spaced wells. A possible explanation lies in laboratory experiments which show that a small increase in water saturation can decrease the gas phase permeability significantly. Conversely, drying out the water saturation during gas cycling in reservoirs or during injection of CO2 into deep saline aquifers affects petrophysical properties such as absolute permeability and capillary pressure. The precipitation of salts from the evaporating brine is one contributor to these effects. In this paper we quantify the effect of small saturations of the wetting phase on nonwetting phase relative permeability. We also show how certain porosity-reducing processes magnify this effect. To compute phase geometry and permeability we use a physically representative network model. The network is extracted from a model rock, built from a dense random packing of spheres modified geometrically to simulate various rock-forming processes. At small saturations (near the drainage endpoint) the wetting phase exists largely in the form of pendular rings held at grain contacts. Pore throats correspond to the constriction between groups of three grains, each pair of which can be in contact. Thus the existence of these pendular rings decreases the void area available for flowing non-wetting phase. Because the hydraulic conductance of the throat varies with the square of the void area, the effect on permeability is disproportionate to the volume occupied by the rings. The same approach quantifies the reduction in permeability by salt precipitation during drying. Convention holds that connate water has little effect on oil or gas permeability because it occupies the smaller pores. Comparing predictions for unconsolidated model rocks with those for cemented model rocks allows one to reconcile this view with the sensitivity reported in the field and the laboratory. We validate the model against experiments, and show that models that do not explicitly account for the phase geometry, such as the Kozeny-Carman equation, cannot capture the observed behavior.