Multi-stage hydraulic fracturing in horizontal wells has proven to be a successful technique for developing unconventional low-permeability oil and gas reservoirs. Despite being vastly implemented by different operators across North America, hydraulic fracture treatments are still not fully comprehended and have proved to have a more complex behavior than initially thought. To fill this knowledge gap, fracture diagnostic tools are used in order to directly or indirectly measure parameters that can provide relevant information to asses a fracture job. Several techniques such as microseismic, tiltmeter mapping, radioactive tracer and production logging have been developed to monitor the fracturing process; however they do not provide accurate quantitative information about fracture characteristics. In this study, we propose to employ dynamic production data as a real-time monitoring tool to characterize hydraulic fractures. A stochastic inverse problem is set up to infer hydraulic fracture characteristics such as fracture conductivity and geometry by integrating fluid production rates.
We start by evaluating the impact that different fracture and reservoir parameters have in the production data. This is done through a sensitivity analysis of multi-stage transverse hydraulic fracturing using a synthetic reservoir model. To estimate hydraulic fracture parameters we employ the ensemble Kalman filter (EnKF). The EnKF is an ensemble based sequential model updating method capable of assimilating production data. The result is a quantitative characterization of hydraulic fractures and automatic history matching. For the latter application, the EnKF also offers several advantages including the ensemble formulation for uncertainty assessment, convenient gradient-free implementation, and the flexibility to incorporate additional monitoring data types.
Examples are presented to illustrate the suitability of the EnKF-based fracture characterization for the inversion of produced oil, gas and water rates to infer fracture geometry and conductivity. This enables a quasi-real-time study of production data as opposed to conventional decline curve analysis practices. Lastly, the application of the EnKF for hydraulic fracture characterization can also be extended by integrating other real-time monitoring data such as DTS data.