Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

SPE Paper 137752

Linking Reservoir Simulators with Fracture Simulators


Carlos Miranda and M.Y. Soliman, SPE, Halliburton; A. Settari, U. of Calgary; and Radim Krampol, TAURUS
Reservoir Solutions Ltd.

Copyright 2010, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Eastern Regional Meeting held in Morgantown, West Virginia, USA, 12–14 October 2010.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Reservoir simulation is currently used as the primary technique to simulate the behavior of almost all types of hydrocarbon
reservoirs. Although several techniques have been developed to simulate the behavior of hydraulic fractures in reservoir
simulators, a lack of accurate modeling of fracture geometry and fracture-fluid leakoff, as well as some other effects on
hydraulic fractures (e.g., non-Darcy flow, dynamic fracture-conductivity behavior, stress-permeability dependence), is
commonly observed in commercial-reservoir simulators.
A different approach is presented in this paper to fill the gap between reservoir simulators and hydraulic fracture
simulators. Software capabilities have been developed to import propped-fracture geometry, proppant-area concentration, and
fracture-fluid leakoff from commercial, grid-based fracture simulators, into a reservoir simulator with capabilities to model
multiphase, non-Darcy flow inside the fracture. This includes the effects of long-term dynamic conductivity, stress-
permeability dependence, and condensates banking, etc. This functionality permits better modeling of fracture-flow behavior
and gives better insight into the cleanup and productivity of fractured wells, which will in-turn allow the user to design better
fractures.

Brief Review of Hydraulic Fracture Modeling in Reservoir Simulation


Hydraulic fractures are modeled in reservoir simulators in several ways, starting from the simple skin-factor method based on
the Cinco-Ley and Samaniego theory (Cinco-Ley and Samaniego 1982), up to more developed methods that automatically
transfer fracture properties from a hydraulic fracture simulator to a reservoir simulator (Shaoul et al. 2007).
One of these approaches implemented techniques that explicitly model the hydraulic fracture by integrating a hydraulic
mechanic model with reservoir two-phase equations and approximating the flow through the invaded zone using piston-like
displacement (Settari 1990). That approach places two independent grid systems (for the hydraulic fracture simulator and
reservoir simulator) together by tracking of proppant fronts, and uses averaging techniques to rigorously simulate the flow
through the fracture and wellbore regions. Although not addressed in the original paper, the technique can also be used in
conjunction with the local grid refinement (LGR) in the reservoir model to improve accuracy. (Stark et al. 2000) showed that
the wellbore can also be rigorously modeled with the fracture.
Another approach developed by Behr et al. (2003) considered the fracture as a series of high-permeability grid blocks,
then the proppant-concentration distribution as well as the fracture-width variation was translated from a fracture simulator.
In that work, the fluid distribution in the invaded zone was estimated using a simplified fracture-propagation model to
reconstruct the development of the fracture, assuming proportional fracture growth, then the leakoff time and leakoff
coefficient were extracted over the fracture to calculate saturation profiles around the fracture using a two-phase flow model.
The above approach was extended to permit access to the internal state variables of the fracture simulator (Shaoul et al.
2007) to initialize the filtrate-fluid saturation around the fracture using the exact leakoff history calculated by the fracture-
simulation model. In that work, a uniform Cartesian grid was overlain on top of the elliptical fracture-conductivity profile
predicted by the fracture simulator, and average conductivity was calculated for each grid in the new Cartesian grid.

New Approach
Although approaches to link hydraulic fracture simulators with reservoir simulators are not new, it is amply accepted that
fractures are too complex to be modeled by simple analytical solutions; therefore, to properly model fracture propagation,
flow mechanisms, and damage effects on fractured wells, a detailed reservoir description is required. In previous approaches,
2 SPE 137752

fracture simulators that make use of simplified analytical and semi-analytical solutions were interfaced with reservoir
simulators.
Taking advantage of robust and sophisticated fracture simulators currently available in the industry, the present approach
proposes to incorporate fracture grid maps that include fracture width, proppant type, and fluid leakoff volume from grid-
based numerical-fracture simulators into a reservoir simulator.
The main differences from previous approaches are:
• Fracture geometry is entirely modeled by a grid-based fracture simulator, which permits the operator to account for
complex fracture growth in multilayered formations or transverse fractures in horizontal wells, eliminating the
assumption of proportional fracture growth used in previous approaches.
• Fracture propagation, fluid leakoff, and proppant transport are numerically solved by the fracture simulators for each
grid cell. The grid map with these properties is then exported to the reservoir simulator, rather than overlaying a
Cartesian grid on top of the elliptical fracture-conductivity profiles, as in previous approaches.
• A reservoir simulator with capabilities to explicitly model hydraulic fractures can be used in this approach
independently if the fracture geometry is provided by an external fracture simulator.
• Fracturing-fluid leakoff is modeled as a different phase in the reservoir simulator with different properties than
formation water.

These capabilities leave room for the reservoir simulator to accurately model several reservoir and fracture effects, like
fracture clean-up, non-Darcy flow, long-term dynamic-conductivity behavior, stress-permeability dependence, and
condensate banking.
Two grid-based fracture simulators commonly used in the industry for fracture-treatment applications in tight-gas and
high-permeability formations were initially linked with a commercially available reservoir simulator, with an option to
incorporate other grid-based fracture simulators. Fig. 1 depicts typical fracture grid maps showing fracture-proppant
concentration from the above-mentioned fracture simulators and corresponding reservoir-simulator fracture gridding.

Fig. 1—Fracture grid maps and corresponding reservoir-simulator fracture grid.

The definition of the same formation top and bottom and identical reservoir properties: permeability, porosity, reservoir
pressure, and water saturation, in both the fracture simulator and the reservoir-simulator models are basic requirements to
properly implement the link. However, the same number of layers is not necessarily required. Grid-based fracture simulators
have the capability to define formation layers directly from log interpretation, creating a considerable number of layers to
account for lithology, stress, and reservoir variations to model complex fracture propagation. Although this same capability
SPE 137752 3

can be used to define the layers in this reservoir simulator, a coarser model in the vertical direction can be defined to save
memory if further complex modeling (e.g., dual porosity/dual permeability, complex fracture network) is planned. Specific
code has been included in the reservoir simulator to read and allocate different grid sizes used by the fracture simulator into
the reservoir model by the means of LGR, as can be seen in Fig. 2.
As mentioned previously, a reservoir simulator with capabilities to explicitly model propped and acid fractures is used in
this approach. This simulator has internal modules to determine fracture conductivity distribution based on effective stress
versus conductivity tables published by proppant manufacturing companies, in a dynamic fashion as a function of time.

Fig. 2—Fracture grid maps inserted into reservoir model and initial conductivity distribution for the two fractures shown in Fig. 1.

Treatment of Leakoff and flowback


The leakoff fluid is included in the reservoir simulator at the time of fracture placement in the grids containing the fracture
and adjacent to them. It is assumed that flow during fracture placement was perpendicular to fracture face (consistent with the
assumptions on leakoff models in fracturing simulators). Then, the fracturing-fluid filtrate (FFF) saturation is increased in the
fracture block and, if necessary, in adjacent blocks, assuming piston-like displacement of the in-situ fluid by FFF, until the
FFF volume associated with the grid is equal to the total leak-off volume in that grid as reported by the fracturing simulator.
However, it is recognized that the in-situ fluid cannot be displaced completely to its residual (immobile) saturation.
Otherwise, the well might not have any initial mobility of the reservoir hydrocarbon. For example, for a water-based FFF in a
gas reservoir initially with conate water at Sw = Swc and residual gas saturation of Sgc, the filtrate will displace the gas with
saturation Sfff = (1-Sgc-Swc)×F where F is a factor slightly less than 1, and Sg will be reduced by Sfff. Then the filtrate volume
of a block with bulk volume V will be Vfff = Vφ Sfff . If this volume is greater than the leakoff volume, Sfff is adjusted. If it is
less, the calculation proceeds in the next block.
This procedure accounts rigorously for the initial volume of filtrate. The simulation then solves both for Sw and Sfff in the
aqueous phase. Therefore, during production, the model can predict both the fracturing-fluid filtrate and in-situ water rates,
which can be history matched to field rates.

Effective Fracture Length and Conductivity


Fracture simulators can only predict created and propped fracture length. Created fracture length is the fracture length
propagated during the treatment, while the propped fracture length is the length supported by proppant after the fracture is
closed. However, for production forecast, reservoir simulators use effective fracture length, which is the length that is open
(i.e. has sufficient conductivity to be contributing to hydrocarbon production after a fracture treatment. Unfortunately, the
most advanced fracture-mapping technologies currently available cannot determine effective fracture length; therefore,
advanced production-data analysis, pressure-transient testing, and/or numerical-reservoir modeling are required to determine
the effective fracture length (Cipolla et al. 2008).
4 SPE 137752

To estimate effective fracture length by matching real production data or to provide different production scenarios for
stochastic analysis, several options are available in the fracture module of the reservoir-simulator package to affect the
fracture geometry imported from the fracture simulators (Fig. 3).
Effective fracture length is frequently smaller than propped fracture length because of the fracture-gel damage and proppant
crushing or embedment (Barree et al. 2003). Instead of shortening the propped length predicted by the fracture simulator
before importing it in the reservoir simulator, one can affect the proppant permeability using reducing factors that depend on
propped width or proppant concentration. This will naturally reduce the effective length for production in areas of poor
proppant placement, as seen on the conductivity maps in Fig. 2. Finally, the incomplete cleanup of fracture fluid FFF at the
tip also reduces the effective length, and this effect is naturally included in our approach.
Fracture conductivity is a dynamic property that decreases with time primarily because of two effects. As the reservoir
pressure depletes the effective stress on proppant pack increases decreasing the proppant permeability. In addition, even at
constant stress, the permeability decreases because of stress corrosion, embedment creep, gel-residue effects, and fines, etc.
To account for this second effect, a table of permeability multipliers versus time can be created to reduce the permeability
with time to match real production data. The resulting conductivity change is a cumulative effect of these and possibly other
factors available in the fracture-interface module of the reservoir simulator (Taurus 2009).

Fig. 3—Fracture module from reservoir simulator.

Simulation Examples
Some examples of typical applications on this approach are discussed in this section. These examples come from real
fracturing treatments in tight-gas and shale-gas formations.

Tight-Gas Formation. Hydraulic fracturing is extensively applied to commercially produce from tight-gas formations
around the world. In this example, a vertical well was completed in Frontier sand at approximately 11,400-ft depth and
fractured using crosslinked fluid and 20/40-mesh proppant sand, following an injection test with KCL brine. The fracture
simulator predicted proppant-concentration distribution from 2.7 lbm/ft2 at the wellbore, to 0.4 lbm/ft2 at a maximum fracture
length of 2,150 ft for this tight-gas formation, with average permeability lower than 0.05 md.
As can be seen in (Fig. 4), the fracture grid map was fully integrated in the reservoir-simulator grid model to capture
detailed fracture geometry predicted by the fracture simulator. Although the same lithology and reservoir properties were
defined in both simulators, the fracture simulator used more refined vertical gridding for accurately capture vertical-fracture
propagation.
LGR of the base grid is performed by default at the tips and around the fracture when it is defined in the grid model;
however, LGR is also used around the fracture to accurately model the characteristic flow regimes (pseudo-linear, bilinear,
and pseudo-radial) observed in fractured wells (Fig. 5). A final, 3-D view can be obtained and used to include digital
structural maps for detailed reservoir description (Fig. 6).
Injected fracturing fluid was treated as a different phase in the fracture simulator with different properties than connate
and formation water. This characteristic permits more accurate modeling of fracture cleanup, which will be discussed in
separate work. In this example, 634 bbl of fracturing-fluid filtrate were recovered in the first year from 2,319 bbl injected
SPE 137752 5

during the fracture treatment (Fig. 7). Very little formation water was produced because it was initially immobile. Although a
proper set of relative-permeability curves for both matrix formation and fracture pack should be incorporated for proper
modeling (Sullivan et al. 2006), this poor water recovery is typical in tight-gas formations, and it has lead to special studies to
better understand this phenomenon (Wills et al. 2009; Wang et al. 2009).

Fig. 4—Fracture grid map fully integrated with reservoir grid model.

Fig. 5—Local grid refinement around the fracture.

Fig. 6—3-D view of hydraulic fracture from the reservoir simulator.


6 SPE 137752

Fig. 7—Production profile after one year, tight-gas example.

A long fracture can be created in this tight-gas formation as a result of massive fracture treatments; however, a short
fraction of the stimulated volume will likely contribute to production because of fracture-face damage and fracture-cleanup
effects. Therefore, a table of fracture-permeability multipliers versus fracture width (Table 1) was used to reduce fracture
conductivity and to shorten fracture length. (Figs. 8 and 9) show the resultant fracture geometry from these corrections and a
corresponding production profile. Again, in this example, little formation water was produced compared to the amount of
fracture-fluid filtrate recovered.
Effective fracture length was reduced to less than 250 ft with fracture-conductivity distribution in the range of 10 to 100
md-ft, which can be considered more representative of the fracture geometry contributing to production in these types of
formations.

TABLE 1—PROPPANT-PERMEABILITY DEPENDENCE ON FRACTURE WIDTH


Fracture Width, in. Perm Multiplier
0.15 0
0.20 0.5
0.25 1

The low water recovery observed after one year in this case (105 bbl) emphasizes the importance of fracture conductivity
for fracture cleanup, as shown in previous studies and in general, and for sustained productivity. Similar results can be
obtained by applying fracture-permeability multipliers versus proppant concentration; then, automatic history matching can
be performed, combining this procedure with stochastic analysis.
SPE 137752 7

Fig. 8—Production profile with modified fracture geometry.

Fig. 9—Modified effective fracture geometry, tight-gas example.

Ductile Shale-Gas Formation. As discussed in recent studies, not all the shale plays have the same properties. They are
heterogeneous, and most of them are different than Barnett shale—a brittle shale-gas formation with isotropic horizontal
stress where complex-fracture networks are believed to be the predominant result from fracture treatments. For that reason,
not all the shale plays should be modeled assuming a complex fracture network. Instead, shale characterization is highly
recommended to determine brittleness and horizontal stress anisotropy, which together will dictate the tendency to generate
planar fractures or complex-fracture networks.
Haynesville has been characterized predominantly as an organic shale ductile formation (Parker et al. 2009) that is under
a normal stress regime with probable horizontal stress anisotropy (LeCompte et al. 2009). Following these conditions, planar
fractures rather than complex fracture networks should be expected in this shale play. However, if natural fractures, fissures,
cleats, and other planes of weaknesses are present, they can produce shear fractures at an angle to the direction of maximum
horizontal stress, adding some complexity around the primary fracture. Also, recently developed special completion
techniques (Soliman et al. 2010) can enhance fracture complexity in horizontal wells drilled in shale-play formations like
Haynesville. Therefore, reservoir modeling in this shale play can be performed to build a composite reservoir model with a
primary dominant fracture and a dual-porosity system covering the stimulated-reservoir volume to account for some fracture
complexity, if present.
8 SPE 137752

In this example, a typical horizontal well was fractured in ten stages using linear gel and crosslinked fluid (hybrid
treatment) and high-strength proppant; (Fig. 11) shows predicted fracture geometry integrated with the reservoir grid model.

Fig. 10—Shalelog analysis, Haynesville Shale, De Soto Parish, Louisiana (Parker et al. 2009).

Fig. 11—Fracture geometry integrated with reservoir model for horizontal well with 10 fractures.
SPE 137752 9

Conclusions
The following conclusions are a result of this work,
• This new approach permits detailed reservoir description and accurate modeling of the physics behind hydraulic
fractures in tight-gas, shale-gas, and high-permeability reservoirs by eliminating simple analytical solutions assumed
in previous approaches.
• Because the fracture propagation, proppant transport, and fluid leakoff have been numerically modeled by the
fracture simulators, the impact of effects like non-Darcy flow, multiphase flow, and stress-dependant permeability
on the productivity of fractured wells can be studied in more detail by the reservoir simulator.
• Either of the two, more-used, commercial grid-based fracture simulators can be linked with the reservoir simulator
to cover a wide range of fracturing applications in tight and high-permeability formations.
• Software capabilities in the reservoir simulator permit proper determination of effective fracture geometry (length
and conductivity) for production matching and sensitivity analysis.

Nomenclature
F Volume factor
Sfff Fracturing-fluid filtrate saturation
Sg Gas saturation
Sgc Residual-gas saturation
Sw Water saturation
Swc Connate-water saturation
V Bulk volume
Vfff Fracturing fluid-filtrate volume
φ Porosity

Acknowledgments
The authors thank Halliburton and Taurus for permitting this work to be published.

References
Barree, R.D., Cox, S.A., Gilbert, J.V., and Dobson, M. 2003. Closing the Gap: Fracture Half Length from Design, Buildup, and Production
Analysis. Paper SPE 84491 presented at the Annual Technical Conference and Exhibition, Denver, Colorado, 5–8 October. doi:
10.2118/84491-MS.
Behr, A., Mtchedlishvili, G., Friedel, T., and Haefner, F., 2003. Consideration of Damaged Zone in Tight Gas Reservoir Model with
Hydraulically Fractured Well. Paper SPE 82298 presented at the European Formation Damage Conference, The Hague, The
Netherlands, 13–14 May. doi: 10.2118/82298-MS.
Cinco-Ley, H. and Samaniego, F., 1982. Transient Pressure Analysis for Fractured Wells. JPT (33): 9 1749–1766. doi: 10.2118/7490-PA.
Cipolla, C.L., Lolon, E. P., and Mayerhofer, M.J. 2008. Resolving Created, Propped, and Effective Hydraulic Fracture Length. Paper IPTC
12147 presented at the International Petroleum Technology Conference, Kuala Lumpur, Malaysia, 3–5 December. doi:
10.2523/12147-MS.
LeCompte, B., Franquet, J.,A., and Jacobi, D., 2009. Evaluation of Haynesville Shale Vertical Well Completions With a Mineralogy Based
Approach to Reservoir Geomechanics. Paper SPE 124227 presented at the Annual Technical Conference and Exhibition, New
Orleans, Louisiana, 4–7 October. doi: 10.2118/124227-MS.
Parker, M., Buller, D., Petre, E., Dreher, D., 2009. Haynesville Shale-Petrophysical Evaluation. Paper SPE 122937 presented at the Rocky
Mountain Petroleum Technology Conference, Denver, Colorado, 14-16 April. doi: 10.2118/122937-MS.
Settari, A., Puchyr, P.J., and Bachman R.C., 1990. Partially Decoupled Modeling of Hydraulic Fracturing Processes. SPE Production
Engineering. (5): 1. 37–44. doi: 10.2118/16031-PA.
Shaoul, J.R., Behr, A., and Mtchedlishvili, G., 2007. Developing a Tool for 3D Reservoir Simulation of Hydraulically Fractured Wells.
SPE Reservoir Evaluation and Engineering Journal.(10): 1 50–59. doi: 10.2118/108321-PA.
Soliman, M.Y. East L., and Augustine J. 2010. Fracturing Design aimed at Enhancing Fracturing Complexity. Paper SPE 130043 presented
at the EUROPEC/EAGE Annual Conference and Exhibition, Barcelona, Spain, 14–17 June. doi: 10.2118/130043-MS.
Stark, A.J., Settari, A., and Jones, J.R. 2000. Analysis of Hydraulic Fracturing of High Permeability Gas Wells to Reduce Non-Darcy Skin
Effects. Journal of Canadian Petroleum Technology. (39): 5 56–63.
Sullivan, R.B., Rushing, J.A., Bachman, R.C., Settari, A., Lujun Ji, Conway, M.W., and Barree, R.D., 2006. Evaluation of Nonlinear
Fracture Relative Permeabilities and Their Impact on Water-Frac Performance in Tight Gas Sands. Paper SPE 98329 presented at the
International Symposium and Exhibition on Formation Damage Control, Lafayette, L.A., 15–17 February. doi: 10.2118/98329-MS.
Taurus, 2009. TFRACI—Taurus Fracture Interface to a Host Reservoir Simulator. User Manual. Ver. 1.4. March. Taurus Reservoir
Solutions Ltd..
Wang, J.Y., Holditch, S.A., and McVay, D.A., 2009. Modeling Fracture Fluid Cleanup in Tight Gas Wells. Paper SPE 19624 presented at
the Hydraulic Fracturing Technology Conference, The Woodlands, Texas, 19–21 January. doi: 10.2118/119624-MS.
Wills, H.A., Miskimins, J.L., and Kazemi, H. 2009. Coupled 3D Numerical Investigation of Hydraulic Fracture Cleanup for Both
Slickwater and Gelled Fluids. Paper SPE 124327 presented at Annual Technical Conference and Exhibition, New Orleans, Louisiana,
4–7 October. doi: 10.2118/124327-MS.

You might also like