Bureau of Economic Geology, The University of Texas at Austin (www.beg.utexas.edu).
SPE Distingushed Lecture Series Luncheon Presentation, February 24, 2004, Austin, Texas
Knowledge of where natural fractures are abundant, open, and capable of transmitting fluid would aid exploration, development, and management of many petroleum reservoirs. Yet this information is usually lacking because conventional analysis of vertical wellbores provides only nonsystematic and incomplete samples of fracture arrays because of the inherent limitations of conventional characterization methods. Thanks to the advent of logging tools that image fractures in the wellbore wall and coring procedures that maintain core integrity in fractured rock, log- and core-based methods usually provide some information on fracture attributes. Yet data are commonly incomplete; the reason is that meaningful samples of fracture networks are inherently difficult to obtain because the probability of intercepting vertical fractures with vertical wells is exceedingly small. For the many areas where large fractures are moderately to widely separate or are arrayed in swarms, even the most complete logging and coring program will frequently miss large fractures. Consequently, a central challenge of subsurface fracture characterization is obtaining data on essential fracture attributes where direct observation is unlikely. Lack of ground truth about fracture attributes prevents predictive models from being tested in a meaningful way and detracts from the effectiveness of fluid flow modeling.
Recent studies show that more systematic and complete data can be acquired by substituting other observations as surrogates for direct fracture measurements. These surrogates are primarily microstructures that do not contribute to fluid flow. Among the parameters that can be accurately assessed using these novel but inexpensive surrogate characterization techniques are fracture orientation, intensity, clustering, and whether large fractures are open or sealed. Tests in exploration and development settings in a wide range of siliciclastic and carbonate reservoirs demonstrates that these methods work and that they can identify fracture targets and quantify parameters essential for accurate fractured reservoir fluid flow modeling when conventional methods fail. Although discrete-fracture modeling provides the most realistic portrayal of fluid flow through fractured rock, this technique is computationally infeasible for simulation of the large volumes in a fractured hydrocarbon field. Dualporosity simulation is the dominant technique for this reason.
We are developing a blended approach to simulation that is both cost effective and grounded on local fracture observation. In conventional reservoir simulations, grid block permeabilities are frequently assigned values larger than those observed in core measurements to obtain reasonable history matches. Even then, accuracy with regard to some aspects of performance such as water or gas cuts, breakthrough times, and sweep efficiencies may be inadequate. In some cases, this could be caused by flow through fractures unaccounted for in simulation. On a bed-by-bed basis, surrogate observations are used to make predictions of key macrofracture attributes. From these predictions, multiple discrete-fracture models are generated for each bed, representing volumes comparable to those of the cells in dual-porosity simulations. Fluid-flow simulations in discrete-fracture models provide a quantitative, observationbased understanding of fluid-flow characteristics and spatial heterogeneity for each bed. These results can then be used to construct a dual-porosity simulation for large regions in the subsurface.
In a test of this approach, a fracture-mechanics-based crack growth simulator, rather than a purely stochastic method, has been used to generate fracture networks with realistic clustering, spacing, and fracture lengths. The surrogate fracture observation was measured subcritical crack index, bed thickness, tectonic strain from microfracture observations, and a microstructure-based surrogate that indicates the presence or absence of open fractures. Coupled fracture-matrix fluid flow simulations of resulting fracture patterns were performed using a finitedifference simulator to obtain equivalent permeabilities that can be used in a coarser scale flow simulation. Results indicate that even though fracture permeability is highly sensitive to fracture aperture, computed equivalent permeabilities are more sensitive to fracture patterns and connectivity. Effects of diagenetic cements completely filling smaller aperture fractures and partially filling larger aperture fractures can have a profound effect on flow.