2005 Salt Lake City Annual Meeting (October 16–19, 2005)

Paper No. 3
Presentation Time: 2:05 PM


FU, Yao1, YANG, Yi-Kun1, FORSTER, Craig B.2 and DEO, Milind1, (1)Department of Chemical Engineering, University of Utah, 50 S Central Campus Dr, Room 3290 MEB, Salt Lake City, UT 84112, (2)College of Architecture+Planning, University of Utah, 375 South 1350 East, Room 235, Salt Lake City, UT 84112, fuyao21@yahoo.com

Reservoir-scale simulations of fault and fracture networks often rely on single-phase averaging of fracture network properties, combined with history-matched estimates of relative permeability (kr) relationships, to define equivalent property values for continuum-type simulators. A new, fully-implicit, 3-D, control-volume finite-element simulator enables three-phase (oil, gas, water) simulations of fluid flow through discrete fracture/fault networks in permeable or impermeable host rock. For example, a faulted reservoir with impermeable matrix contains an orthogonal network of 29 discrete, vertical features (with uniform width of 0.03 m) grouped into two sets. The network covers an area 305 m by 311 m with uniform feature height of 65 m and a total feature volume of 1,309 m3. Spacing, orientation and trace length vary within each set of features. One injection well and one production well each pierce features at opposite corners of the model domain. Uncertainty in kr relationships for fault-affected rocks is represented by two end-member kr relationships: (1) non-linear relationships typical of porous media, and (2) linear relationships typical of fractures. Assigning uniform kr properties and absolute k of 1,000 md in all features yields cumulative production in the linear kr case (0.6 OOIP) that, after 800 days, is 27% greater than that of the non-linear kr case. Assigning a distribution of kr curves throughout the fracture network yields intermediate results. Most unswept oil is trapped in parts of the network that are weakly connected to injection and production wells. Extending one feature by 10s of meters, however, significantly increases total production and illustrates the large impact of uncertainty in estimating network connectivity. Gas released during depressurization rises to the top of the features while injected water flows along the bottom of the features. Because wells cannot be perforated through the full height of a fault zone, early water breakthrough occurs if the production well pierces a permeable feature near the bottom of the network. A theoretical history matching exercise yields several equivalent kr and absolute k combinations (for a region embedded between the injection and production wells) that mimic the discrete-feature production results with varying degrees of success.