Development of Proxy Model for Hydraulic Fracturing and Seismic Wave Propagation Processes

Abstract

Characterization of discrete fracture networks is necessary for unconventional reservoir development, as they control the flow of fluids toward the hydraulically fractured well. Interpretation of microseismic data provides information about the discrete fracture network in the vicinity of a well. While microseismic interpretation is currently based on plotting the swarm of microseismic events, the inference of fracture-related information from such data is likely to be non-unique. To address this non-uniqueness, the presented workflow involves applying a forward model that produces a synthetic seismogram corresponding to a suite of natural fractured reservoir models. The characteristics of the generated seismograms can be compared with the observed seismogram to determine the suite of fracture models that best reflect the observed seismic signature. The available full physics models are computationally expensive and cannot be applied within this framework. Hence, a proxy model is developed that is computationally inexpensive so that it can be applied on a large ensemble of models within the Bayesian model selection framework described above. Seismic wave propagation during a fracturing job involves many intermediate processes such as diffraction and reflection. As analytical solutions for most of these processes exist, a coupled analytical model is proposed. Firstly, a hydraulic fracture propagation model is coupled with a model for the interaction between a hydraulic fracture and a natural fracture. This interaction results in slip events along natural fractures that in turn can trigger a seismic event. With knowledge of the location of the slip event, a Green’s function solution is applied to model the propagation of the seismic wave. Using the results from the slip event and Green’s function, a synthetic seismogram is generated. Finally, the seismic signature from multiple slip events along several natural fractures is combined. Validation of individual elements of the coupled proxy model to experiments is shown. A comparison of the developed proxy model with more computationally expensive models is also performed.