The large uncertainty in fracture characterization for shale gas reservoirs seriously affects the confidence in making forecasts, fracturing design, and taking recovery enhancement measures. It is of significance to make the most of multiple data in fracture characterization, including the measurements of hydraulic fracturing treatment, core, microseismic, production data, etc. Besides, handling the complex fracture networks (CFNs) explicitly is demanded in large-scale field case studies. This paper presents an efficient method to characterize the complex fracture networks (CFNs) and reduce the uncertainties by integrating stochastic CFNs modeling, reservoir simulation, and assisted history matching. In CFNs modeling, the geometry of CFNs is generated stochastically constrained by the measurements of hydraulic fracturing treatment, core, and microseismic data, and a stochastic parameterization model is used to generate an ensemble of initial realizations of the stress-dependent fracture conductivities. To efficiently simulate the discrete fractures in field cases, the edge-based Green Element Method (eGEM) is modified by applying a steady-state fundamental solution and the technique of local grid refinement (LGR). Assisted-history-matching based on EnKF is implemented to calibrate the CFNs models and further quantify the uncertainties. The proposed method is applied to a field case from a multi-fractured horizontal shale gas well. The results show that the computation of the modified eGEM is much cheaper than the original eGEM when the steady-state fundamental solution and the technique of LGR are applied. Besides, acceptable precision can be obtained with relative coarse grids, for example, 50 m primary grids and 25 m refined grids near the fractures can be used in the test case. These advantages make the modified eGEM quite efficient in assisted history matching of large-scale field applications. The history matching results and EUR distributions show that the uncertainties are significantly reduced after 5 steps assimilation, and good results can be obtained after 10 steps assimilation. The distribution ranges of the matrix permeability and porosity are significantly narrowed down after history matching, while the matrix compressibility is hard to be determined because the rock compressibility is so small comparing to gas compressibility that gas production is not sensitive to it. The results also show that only part of the fractures are well-propped because only the fractures in a certain range of the wellbore are with good properties, that is high initial fracture conductivities and low conductivity moduli. • An integrated method for complex fracture networks (CFNs) characterization. • Geological, seismic, fracturing and production data can be assimilated. • An efficient froward model for field-scale reservoir simulation with CFNs. • Property calibration and uncertainty quantification of CFNs.
Read full abstract