This work aims to address a challenge posed by recent observations of tightly-spaced hydraulic fractures in core samples from the Hydraulic Fracturing Test Site (HFTS). Many fractures in retrieved cores have sub-foot spacing, which is at odds with conventional models where usually one fracture is initiated per cluster. Since it is unrealistic to explicitly model all densely-spaced fractures, we develop a new upscaling law that enables existing simulation tools to predict reservoir responses to fracture swarms. The upscaling law is derived based on an energy argument and validated through multiscale simulations using a high-fidelity code, GEOS. The swarming fractures are first modeled with a spacing that is much smaller than the cluster spacing; these fractures are then approximated by an upscaled, single fracture based on the proposed upscaling law. The upscaled fracture is shown to successfully match the energy input rate and produce the total fracture aperture and average propagation length of the explicitly simulated swarm. Afterwards, the upscaling approach is further implemented in 3D field-scale simulations and validated against the HFTS microseismic data of a horizontal well in the Middle Wolfcamp Formation. Our results show that hydraulic fracture swarming can significantly affect fracture propagation behaviors compared with the propagation of single fractures as assumed by conventional modeling approaches. Under the considered situations, the conventional case entails fast propagation speed that far exceeds that indicated by the microseismic data. We also illustrate this discrepancy can be reduced readily through the implementation of the upscaling law. Our results demonstrate the importance of accounting for the fracture swarming effect in field-scale simulations and the efficacy of this approach to enable realistic predictions of reservoir responses to fracture swarms, without explicit modeling all tightly-spaced fractures observed in the field.

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