Data assimilation for fractured shale gas reservoirs using Ensemble Kalman Filter.

摘要

Production of shale gas reservoirs depends on natural and hydraulic fractures, which represent a significant challenge in numerical simulation. Unknown fracture characteristics such as location, orientation, aperture, and conductivity make reservoir modeling difficult. Even by knowing these properties, numerical models must be refined to capture the complex flow behavior around the fractures. Discrete fracture networks require model dependent unstructured gridding. Furthermore, for history matching and data assimilation, fracture characteristics must be updated, which causes changing the entire gridding of the model and is complicated and time consuming. Systematic history matching of shale gas reservoirs has not yet been addressed in the literature. In this study, we use shale gas wells' measurements to estimate the fractured reservoirs properties using Ensemble Kalman Filter (EnKF), a minimum mean square error data assimilation tool. We propose using dual porosity dual permeability modeling (DPDP), an averaging technique that does not require the knowledge of the fracture network characteristics. In the combined EnKF/DPDP methodology, numerical models are updated without changing the gridding as more measurements become available. We introduce and develop a new DPDP compartmentalized modeling, which represents the complex fracture network around a well. The updated models reproduce the historical performance of the reservoir and predict its future behavior.;We test our proposed methodology on synthetic and real field cases from Appalachian Marcellus shale. It is shown that the algorithm does not require information about the locations and orientations of the fractures. We also show that if the knowledge about the fracture network statistics is available, it can be integrated into the algorithm yielding more accurate estimates of the reservoirs' field properties such as fracture porosity and permeability. Gridding is simple and DPDP models are simulated much faster than either the refined or DFN models.;It is illustrated that the proposed methodologies provide a reliable and robust data assimilation and modeling tool for history matching of fractured reservoirs. They do not require changes in gridding and take less CPU time. Although the proposed methods are applied to shale gas reservoirs in this dissertation, their application can be extended to other types of fractured reservoirs.