In Imperial Oil’s Cold Lake field, Cyclic Steam Stimulation (CSS) is used to recover bitumen from the Clearwater formation. At initial reservoir conditions, fluids are essentially immobile due to high bitumen viscosity and due to very low relative permeability to water. The injection of steam is achieved by injecting at pressures that induce hydraulic fractures in the formation. Numerical simulation techniques typically handle fracturing by imposing a large increase in permeability as the pressure exceeds a fracture pressure.
Standard simulation approaches using an ExxonMobil in-house simulatori to modeling fracturing require an excessive pressure gradient in order to propagate a fracture. As a result, the pressure of the fracture blocks is overestimated, resulting in an unrealistic overestimate of leak-off from the fracture. The fracture area is thereby significantly underestimated and possible fluid communication between wells is not properly modeled.
The term "dynamic fracturing" refers to an approach that improves hydraulic fracture modeling by using a variation of upstream weighting to calculate the effective permeability between fracture blocks. The upstream weighting is done indirectly using changes in void ratio, which is a measure of pore dilation due to pressure change. The method has been calibrated with field data from two Cold Lake pads.
Early attempts to use this method encountered numerical stability problems. This resulted in very slow run times, and in many instances, cases could not be completed. The method has been improved using a simple dampening scheme making it practical to use in multi-well models. This approach is aimed at improving predictions of fluid distribution in early cycles of CSS including predictions for infill well performance. More realistic fluid communication modeling improves subsequent infill performance predictions. Useful comparisons of performance from different steam strategies require good predictions of fluid distribution caused by each strategy. This work has imporved our ability to model the CSS process and to predict field behaviour.