This paper was prepared for the 2009 SPE/EAGE Reservoir Characterization and Simulation Conference.
Well log and core information, seismic surveys, outcrop studies, and pressure transient tests are usually insufficient to generate representative 3-D fracture network maps individually. Any combination of these sources of data could potentially be used for accurate preparation of static models.
Our previous attempts showed that there exists a strong correlation between the statistical and fractal parameters of 2-D fracture networks and their permeability (Jafari and Babadagli, 2009). We extend this work to fracture network permeability estimation using the statistical and fractal properties data conditioned to well test information. For this purpose, 3-D fracture models of nineteen natural fracture patterns with all known fracture network parameters were generated initially. It is assumed that 2-D fracture traces on the top of these models and also 1-D data from imaginary wells which penetrated the whole thickness of the cubic models were available, as well as pressure transient tests of different kinds. The 1- and 2-D data include statistical parameters (density and length distribution) and ten different fractal characteristics of different properties of the fracture system. Next, the permeability of each 3-D fracture network model was calculated and then converted into a grid based permeability map for drawdown well test simulations using commercial software packages.
Finally, an extensive multivariable regression analysis using the statistical and fractal properties and well test permeability as independent variables was performed to obtain a correlation for equivalent fracture network permeability. The equation was validated against different natural and synthetic fracture network patterns. The cases requiring expensive well (logsand cores) and reservoir (pressure transient tests) data were identified. This approach and correlation is expected to be a useful tool for practitioners asit reduces the computational time in static model preparation significantly and utilizes the available data effectively.
The two most common approaches proposed for model transport in fractured reservoirs (dynamic modeling) are single and dual-porosity models. These methods require grid based representation of fracture network properties like porosity and permeability. The discrete fracture network approach is morecapable of representing the complex nature of fracture networks but they are limited in modeling complex dynamic processes. Hence, an accuraterepresentation of a fracture network and its equivalent permeabilitydistribution is a crucial task in dynamic modeling.
