This paper describes the development and capabilities of a novel and unique
tool that interfaces a hydraulic fracture model and a reservoir simulator. This
new tool is another step in improving both the efficiency and consistency of
connecting hydraulic fracture engineering and reservoir engineering.
The typical way to model hydraulically fractured wells in 3D reservoir
simulators is to approximate the fracture behavior with a modified skin or
productivity index (PI). Neither method captures all the important physics of
flow into and through the fracture. This becomes even more critical in cases of
multiphase flow and multilayered reservoirs. Modeling the cleanup phase
following hydraulic fracture treatments can be very important in tight gas
reservoirs, and this also requires a more detailed simulation of the fracture.
Realistic modeling of horizontal wells with multiple hydraulic fractures is
another capability that is needed in the industry. This capability requires
more than an approximate description of the fracture(s) in the
To achieve all the capabilities mentioned above, a new tool was developed
within a commercial lumped 3D fracture-simulation model. This new tool enables
significantly more accurate prediction of post-fracture performance with a
commercial reservoir simulator. The automatically generated reservoir simulator
input files represent the geometry and hydraulic properties of the reservoir,
the fracture, the damaged zone around the fracture, and the initial pressure
and filtrate fluid distribution in the reservoir. Consistency with the
fracture-simulation inputs and outputs is assured because the software
automatically transfers the information.
High-permeability gridblocks that capture the 2D variation of the fracture
conductivity within the reservoir simulator input files represent the fracture.
If the fracture width used in the reservoir model is larger than the actual
fracture width, the permeability and porosity of the fracture blocks are
reduced to maintain the transmissibility and porous volume of the actual
fracture. Both proppant and acid fracturing are handled with this approach. To
capture the changes in fracture conductivity over time as the bottomhole
flowing pressure (BHFP) changes, the pressure-dependent behavior of the
fracture is passed to the reservoir simulator.
Local grid refinement (LGR) is used in the region of the wellbore and the
fracture tip, as well as in the blocks adjacent to the fracture plane. Using
small gridblocks adjacent to the fracture plane is needed for an adequate
representation of the filtrate-invaded zone using the leakoff depth
distribution provided by the fracture simulator.
The reservoir simulator input can be created for multiphase fluid systems
with multiple layers and different permeabilities. In addition, different
capillary pressure and relative permeability saturation functions for each
layer are allowed.
Historically, there have been three basic approaches commonly used for
predicting the production from hydraulically fractured wells. First, analytic
solutions were most commonly used, based on an infinite-conductivity or, later,
a finite-conductivity fracture with a given half-length. This approach also was
extended to cover horizontal multiple fractured wells (Basquet et al. 1999).
With the development of reservoir simulators, two other approaches were
For complicated multiwell, multilayer, multiphase simulations (i.e.,
full-field models), the fracture stimulation was usually approximated as a
negative skin. This is the same as increasing the effective wellbore radius in
the simulation model. An alternate approach, developed initially for tight gas
applications, was to develop a special-purpose numeric reservoir simulator that
could explicitly model the flow in the fracture and take into account the
special properties of the proppant, such as the stress-dependent permeability
or the possibility of non-Darcy flow. Such models typically were limited to a
single-layer, single-phase (oil or gas) situation.
© 2007. Society of Petroleum Engineers
View full textPDF
- Original manuscript received:
14 July 2005
- Meeting paper published:
21 November 2005
- Revised manuscript received:
25 August 2006
- Manuscript approved:
10 October 2006
- Version of record:
20 February 2007