Abstract Fractures are widely distributed in the upper crust, which is important to transport processes in groundwater, geothermal, and hydrocarbon reservoirs. Little attention has been paid to the heterogeneity and accuracy of upscaled equivalent fracture models (EFMs) regarding the complex fracture geometries during numerical reservoir simulation. This paper investigates how fracture geometric properties affect the accuracy of highly heterogeneous EFMs. An ensemble of synthetic fractured porous rocks is generated stochastically with varied fracture length, fracture density, and aperture-length correlation. Each realization is upscaled to EFM with the recently developed multiple boundary method. Flow problems are solved for EFMs and discrete fracture models (DFMs) simultaneously, and their flow rates are compared. The results demonstrate that the heterogeneity of EFMs decreases with the number and length of fractures and increases with the aperture-length correlation component. The flow rates calculated by EFMs fit well with those by DFMs with a slope around 1.08 and R2 coefficient above 0.98. The accuracy of the upscaled EFM decreases when the aperture-length correlation changes from sublinear to linear. The uncertainty of upscaling accuracy decreases with dimensionless fracture density. The findings of this study can help for a better understanding of the effect of the discrete fracture geometries on the performance of upscaled EFMs and choosing the appropriate models for different kinds of fractured reservoirs.
Read full abstract