Microseismicity is a physical phenomenon which allows us to estimate the production capability of the well after hydraulic fracturing (HF) in a naturally fractured (NF) reservoir. Some of the microseismic events are reactivations of NFs induced by a direct hit of HF, while others are induced by the fluid leak-off from the previous stages or by elastic waves emitted into the reservoir with hydraulic fracture plane propagation. The former NFs have a chance to be propped there as the latter will not significantly increase their contribution to the production. Identification of such microseismic events helps to reduce uncertainty in the description of fracture network geometry.

Based on inferred data from core analysis NF densities and orientations, we generated multiple realizations of the semi-stochastic Discrete Fracture Network (DFN). In order to constrain them, we used time evolution of microseismic cloud in addition to results of core analysis. Fluid and proppant pumping schedule is used to identify such microseismic events because they should be located close to the pressure diffusion front generated by hydraulic fluid. Events outside of proposed region may be triggered by other factors, such as stress-strain relaxation from other stages and correspondent fractures. In most cases, they are not wide enough to take proppant from the main HF. This approach was used to reduce range of production for DFN realizations.

This workflow is implanted to a 15-stage hydraulic fracture treatment on a horizontal well placed in a siltstone reservoir with intrinsic fractures. The spatio-temporal dynamics of microseismic events are classified into two groups by the front of nonlinear pressure diffusion caused by 3-dimensional hydraulic fracturing, considered as effective and ineffective events. DFNs with only effective microseismicity and with all the induced events are generated. Then, two types of DFN related uncertainties on production are performed to evaluate the impact of filtration. Results of aleatory uncertainty quantification caused by the randomness of DFN modeling indicate the filtered events can generate a production DFN with a more consistent connected fracture area. Moreover, sensitivity analysis caused by lack of accuracy in natural fracture characterization shows the production area of DFN with filtration process is more insensitive to the variation of fracture parameters. Finally, a history match with production data and pressure data indicates this DFN model properly represents the reservoir and completion.

Our methodology characterizes well the conductive fracture network utilizing core data, microseismic data, and pumping schedule. It could restore the true productivity of each fractured stage from a massive microseismic cloud, which helps understand the contribution of fracturing job right after the treatment.

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