Вы находитесь на странице: 1из 36

SPE 140514

A Discrete Fracture Network Model for Hydraulically Induced Fractures:


Theory, Parametric and Case Studies
Bruce R. Meyer, SPE, Meyer & Associates, Inc. and Lucas W. Bazan, SPE, Bazan Consulting, Inc.

Copyright 2011, Society of Petroleum Engineers


This paper was prepared for presentation at the SPE Hydraulic Fracturing Technology Conference and Exhibition held in The Woodlands, Texas, USA, 24-26 January 2011.
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 mus t contain conspicuous acknowledgment of SPE copyright.

Abstract
A solution methodology and mathematical formulation for an induced hydraulic Discrete Fracture Network (DFN) numerical
simulator is presented. Although most conventional fracture treatments result in bi-wing fractures, there are naturally
fractured formations that provide geomechanical conditions that enable hydraulically induced discrete fractures to initiate and
propagate both vertical and horizontal fractures in the three principal planes. The fundamental first-order DFN continuity,
mass, momentum, and constitutive equations are developed and formulated for a pseudo-three-dimensional (P3D)
hydraulically induced fracture system. The theoretical foundation and concepts for multiple, cluster, complex, and discrete
fracture network growth are presented.
Discrete fracture interactions as a result of fluid loss and mechanical interference are discussed and included in the
modeling. A new extended wellbore pressure loss and storage concept in the fracture mid-field is introduced. The Extended
Wellbore Storage (EWS) region in the fracture mid-field accounts for the high fracture pressures observed when fracturing in
horizontal or highly deviated wellbores and the associated steep pressure decline during closure. Numerous proppant transport
scenarios are formulated and presented for the transport of proppant in the dominant and secondary fracture network system.
Application of this technology will provide the operator with a systematic approach for designing, analyzing, and
optimizing multi-stage/multi-cluster transverse DFN induced hydraulic fractures in horizontal wellbores. This paper provides
the foundation for predicting the propagation and behavior of discrete fracture networks in shales and unconventional
formations along with the associated generated Stimulated Reservoir Volume (SRV). Numerous parametric and case studies
are provided illustrating the technology and engineering application of the DFN modeling.
Introduction
Hydraulic fracturing and horizontal drilling are the two key technologies that have made the development of shale formations
commercially economical. Hydraulic fracturing has been the major and relatively inexpensive stimulation method used for
enhanced oil and gas recovery in the petroleum industry since 1949. The multi-stage and multi-cluster per stage fracture
treatments in horizontal wellbores create a large stimulated reservoir volume (SRV) (see Mayerhofer et al. (2008)) that
increases both production and estimated ultimate recovery (EUR).
As stated above, most conventional fracture treatments result in bi-wing fractures. However, some naturally fractured
coal and shale formations have geomechanical properties that allow hydraulically induced discrete fractures to initiate,
propagate and create complex fracture networks. The microseismic data collected during a fracture treatment can be a very
useful diagnostic tool to calibrate the fracture model by inferring DFN areal extent, fracture height and half-length. Pressure
history matching of the fracture treatment and production analyses are additional diagnostic procedures the engineer can use as
assurance of the created DFN and SRV.
Davidson et al. (1993) presented detailed minifrac evaluation results for the Gas Research Institutes (GRI) fourth Staged
Field Experiment (SFE 4) conducted in the Frontier formation of southwest Wyoming. This paper presented a discussion on
the possibility of multiple hydraulic fractures being created in formations that contain natural fractures, including numerous
references cited in the literature identifying the existence of multiple fractures created during the hydraulic fracturing process.
The authors presented scenarios whereby multiple fractures could be initiated from a vertical wellbore, including: 1) each
fracture could be propagating from the wellbore originating from a different set of perforations or 2) one main fracture may be
extended from the wellbore and a secondary fracture may split off, forming a fracture spray. Their paper also presented an
analysis of abnormally high fracture treating pressures caused by complex fracture growth.
Weng (1993) was one of the first to present results of an analytical study of hydraulic fracture initiation and propagation

SPE 140514

from deviated wellbores. Weng discussed the interaction and link-up of starter fractures from the wellbore, the fracture
turning and twisting, and finally the tortuous paths created before fracture re-orientation and creation of multiple fractures. He
presented schematics depicting mechanisms for fractures initiating from perforations to link up by rotating their leading edges
until they were parallel to the wellbore. Weng hypothesized that the initial high pressures in deviated wellbores are probably
caused by multiple fractures or tortuosity in the near well region. Weng also stated that high fluid pressures will always help
fracture link-up. A number of complex fracture schematics were also presented. Hainey and Weng (1995), in a follow up
paper, discussed the migration of multiple fractures from deviated wellbores and presented an analysis of multiple fractures
including mechanical (stiffness) interaction. They concluded in order to minimize the adverse effect of multiple fractures,
one needs to minimize the number of fractures propagating from the wellbore.
King et al. (2008) presented a paper titled Increasing Fracture Path Complexity and Controlling Downward Fracture
Growth in the Barnett Shale. King states that The use of downhole microseismic during fracturing has defined sound
patterns that confirm the fracture flow path in these stimulated shale wells is actually linking intersecting natural fractures
into a network of fractures with a flow path width of several hundred feet. These flow paths often have patterns of fracture
growth at right angles to the primary fracture direction. Numerous examples of these right angle secondary fractures that
define the complex fracture network were presented. King further states that The most highly developed complex or network
fracture development is possible where natural fracture systems follow a butt and cleat system common in coals and
shale and the presence of intersecting secondary fractures opens the possibility of strong fracture complexity
development. King also states that The development of both primary and secondary fractures is possible when the
maximum and minimum stresses are similar.
Warpinski et al. (2008) and Cipolla et al. (2008) presented papers on fracture treatment issues related to fracture
complexity. The authors discussed various fracture growth and complexity scenarios which they divided into four groups: 1)
planar-coupled growth, 2) planar de-coupled growth or fissure opening, 3) complex growth (communicating and noncommunicating) and 4) network growth. They also discussed the importance of microseismic mapping and how it could infer
large-scale fracture complexity, as well as the fact that fracture pressure analysis could be used to determine small-scale
complexity. Figures 1a and 1b illustrate complex fracture scenarios presented by Cipolla et al. (2008), Warpinski et al. (2008)
and Mayerhofer et al. (2008).

Figure 1a Complex fracture growth and complex scenarios.

Figure 1b Complex fracture growth and complex scenarios.

(SPE 119890)

(SPE 115769)

Cipolla et al. (2008) also proposed three limiting scenarios for proppant placement in complex fractures as shown in
Figure 2:
Case 1: The proppant is evenly distributed throughout the complex fracture system.
Case 2: The proppant is concentrated in a dominant planar fracture with an unpropped complex fracture system
accepting fluid only.
Case 3: The proppant settles and forms pillars that are evenly distributed within the complex fracture system.

SPE 140514

Figure 2 Proppant transport scenarios. (SPE 115769)


Yeager et al. (2010) presented a paper on the importance of injection/fall-off tests using the Nolte G function and after
closure analysis (ACA) to determine the ISIP, closure stress, net pressure, fluid efficiency, reservoir pressure, and formation
permeability for horizontal Marcellus wells. Yeager referenced the work of Engelder, Lash, and Uzcategui (2009) whereby
they performed extensive mapping of natural fracture patterns on Marcellus shale outcrops throughout the play. Their results
confirmed the existence of two or more joint sets within the formation. The primary set of natural fractures ( J1 ) is within a
few degrees of the maximum horizontal stress in accordance with the world stress map (S Hmax) (Zoback (1992)). A second set
of natural fractures ( J 2 ) was identified in Marcellus outcrops. Engelder et al. concluded that the J 2 set was consistently
orthogonal to the J1 set (see Figure 3d). Yeager also confirmed that the pressure decline was not influenced by pressure
dependent leak-off or height recession after closure, but rather by two fracture closure pressures that may be a confirmation
of the existence of J1 and J 2 natural fracture sets and extended wellbore effects. The difference between Pc1 and Pc2 may be
indicative of the anisotropy of SHmax ( 2 ) and Shmin ( 3 ) or the anisotropy of the vertical and minimum horizontal stresses (see
Jacot et al. (2010)). Yeager concluded that Multiple closure pressures were observed while performing minifrac analyses in
two of the three horizontal wells. We believe these multiple closures are indicative of complex fracture network behavior as a
result of stress anisotropy, creation of complex fractures in the extended wellbore region, and hydraulically induced secondary
fractures perpendicular to the maximum horizontal stress. Fontaine et al. (2008) discussed the concept of multiple competing
fractures initiating from separate perforations and discussed the complex nature of the Marcellus formation. Fontaine also
made reference to the extreme layering and the existence of natural fracturing and/or jointing within the Marcellus.
Dahi-Taleghani and Olson (2009) in their opening remarks state: Recent examples of hydraulic fracturing diagnostic data
suggest complex, multi-stranded hydraulic fractures geometry is a common occurrence. This reality is in stark contrast to the
industry-standard design models based on the assumption of symmetric, planar, bi-wing geometry. The interaction between
pre-existing natural fractures and the advancing hydraulic fracture is a key condition leading to complex patterns. DahiTaleghani and Olson then present a detailed numerical study using a finite element code to model the interaction between a
growing hydraulic fracture and the surrounding natural fractures. Their studies indicate that the effect of stress anisotropy
could magnify the effect of natural fractures on hydraulic fracturing propagation.
Jacot et al. (2010) and Bazan et al. (2010) presented papers on technology integration in the Marcellus and Eagle Ford
shales as a methodology of integrating minifrac analysis, hydraulic fracturing (DFN), diagnostic flowmeter logs and
microseismic technologies with the production response for multiple transverse vertical fractures in horizontal wells with the
aim of improving their stimulation program to enhance gas production. Both authors demonstrated that microseismic data
collected during a fracture treatment can be a very useful diagnostic tool to calibrate the fracture model by inferring DFN areal
extent, fracture height and half-length and fracture orientation.
Xu et al. (2009) presented a hydraulic fracture network growth model. Their growing elliptic shale formation consisting
of two perpendicular sets of equally spaced vertical planar open fractures, respectively parallel to the direction of the two
horizontal principal stresses, is used to provide a mathematically equivalent representation of a hydraulically induced fracture
network. They further state that their fracture growth model fully accounts for mechanical interactions between the
fracturing fluid and fracture walls and among nearby fractures. Xus paper also provided a case study from the Barnett Shale
with microseismic mapping that illustrated complex fracture behavior and a comparison with their model results.
The original concept for the discrete fracture network model presented in this paper was formulated in 2008 based on the
evidence of microseismic mapping and a realization that, for naturally fractured formations, induced fractures may not all be
bi-winged. Olson (2008-2010) provided an extensive discussion and photo assembly of formation outcrops. The photos as
shown in Figures 3a-3d and papers by Olson (1995, 2003, 2004) helped inspired this DFN formulation. Figure 3a shows a

SPE 140514

photo of an orthogonal fracture outcrop in a sandstone formation. Figure 3b shows an example of saturated fracture spacing
from the Oil Mountain area in Wyoming. Olson explains that This is an unfolded outcrop near Oil Mountain, which is a
thrust-cored anticline. Here the fractures running up and down in the diagram which parallel the fold axis of the nearby
anticline, have regular spacing that is about equal to the bed thickness for the formation (the Frontier Sandstone). Figure 3c
illustrates a cliff outcrop with joint spacings approximately proportional to bed thickness. Figure 3d shows an outcrop of the
Marcellus shale in a creek in Leroy New York (Engelder and Lash (2008)).

Figure 3a Orthogonal Fractures in Sandstone

Figure 3b Saturated Strike Joints from the Oil Mountain Area

(Jon Olson (2008-2010))

(Jon Olson (2008-2010)).

Figure 3c Joint Spacing Proportional to Bed Thickness

Figure 3d Marcellus Outcrop (Engelder and Lash (2008)).

(Jon Olson (2008-2010)).

Adda-Bedia and Amar (2000) presented a paper titled Fracture Spacing in Layered Materials and stated that Parallel
open mode brittle fractures, or joints, in layers are common structures in nature. They also stated that the crack spacing to
layer thickness cannot decrease below a certain threshold, and using results from their model indicated that the spacing
threshold is of order unity. Sagy and Reches (2004) discussed the joint density in layered rocks with reference to many field
studies indicating that the mean or medium joint spacing in layered rocks [is] often linearly related to layer thickness. Their
model suggests that the saturation intensity ( Ds h / d ) for realistic geological cases converges to the range of
Ds 0.75 3.0 (i.e., d / h 1 3 4 3 ). Tang et al. (2008) also discussed opening mode fractures in layered material and
state that the fracture spacing scaled to the thickness of the fractured layer. Numerically they showed that the critical fracture
spacing to layer thickness ratio varied from about 1.0 to 2.0 as a function of the homogeneity index, ( m 1.1 15 , where a
larger m implies a more homogeneous material).
Peidro (2009) presented a literature review with finite element simulations of the relationship between fracture density and
the mechanical layering of rocks. His literature review is extensive with numerous studies indicating that there is a linear
relationship between joint spacing and layer thickness. Numerically Peidro concluded Only Youngs moduli of a layer and
its thickness have an influence in the propagation of a fracture or set of fractures. Thus, the intensity of propagation will
decrease with the decreasing stiffness, and also with the increasing thickness of the incompetent layer. This of course
describes the compliance characteristics of the PKN model.
Meyer & Associates, Inc. (2009) released the first commercially available discrete fracture network simulator to the

SPE 140514

petroleum industry in April of 2009. The discrete fracture network model is based on a methodology similar to that presented
by Warren and Root (1963) for dual porosity naturally fractured reservoirs. Figure A.1 illustrates the idealization of a
heterogeneous porous media and depicts an independent system of secondary porosity [that] is superimposed on the primary
or intergranular system. The DFN model we present in this paper accounts for discretized P3D fracture growth in the three
principal planes and includes options for proppant transport distribution throughout the primary and secondary fracture
network. Detailed descriptions of the DFN model formulation and theory, and parametric and case studies are presented.
Theory
This section summarizes the governing equations describing the numerical solution for a discrete fracture network simulator.
Details of the DFN formulation are given in the Appendices. Appendix A summarizes the solution methodology and major
assumptions for the DFN model. Appendix B introduces a new extended wellbore storage (EWS) and dissipation function that
accounts for the complex fracture behavior and pressure decline in the extended or mid-field region of the fracture.
The discrete fracture network may be composed of secondary fractures in all three principal planes. The dominant or
primary fractures propagate in the x-z plane perpendicular to the minimum horizontal stress, . The y-z and x-y plane
fractures propagate perpendicular to the and , stresses, respectively. The discrete fractures created in the x-z (major axis)
and y-z (minor axis) planes are vertical, while the induced fractures created in the x-y plane are horizontal. The spacing in the
x-z, y-z, and x-y planes are y , x , and z , respectively. Figures A.2 through A.4 illustrate the typical fracture profiles
generated for an induced DFN in the principal planes.
The fundamental Discrete Fracture Network (DFN) solution methodology is based on satisfying continuity, mass
conservation, constitutive relationships and momentum equations. These equations are then solved numerically. To illustrate
the methodology, a set of equations are presented in Appendix A for flow in an ellipsoidal slot. Although the 3D equations are
more involved, the referenced derivation provides the reader with a general understanding of the principles used in the DFN
solution. The formulation uses a 3D ellipsoidal gridded approach and includes fracture interaction (fluid loss and mechanical),
proppant transport, and planar fracture propagation in the principal planes. Numerous complex, DFN, and proppant transport
options are available.
Mass Conservation
The governing mass conservation equation, for an incompressible slurry in the fracture, states that the volume of slurry
injected into the fracture minus the volume loss by leakoff, Vl (t ) , and spurt loss, Vsp (t ) , must be equal to the fracture volume,
V f (t ) , which from Eq. A.1 is

t
0

q( ) d Vl (t ) Vsp (t ) V f (t )

(1)

The DFN fluid loss due to leakoff and spurt loss for N discrete fractures of surface area A from Eq. A.21 is
N

i 1

Vl (t ) 2

A
0

v c da d 2 s p c da
i 1

(2)

where the fluid loss multiplier l is defined as:

l ( A / ADFN 1)l 1

(3)

and the total discrete fracture area is given by ADFN Ai .


i 1

The DFN geometric characteristics are summarized in Eq. A.38 and given below
N

LDFN L j
x , y , z j 1

ADFN A j
x , y , z j 1

VDFN V j
x , y , z j 1

wDFN VDFN ADFN

(4)

Continuity Equation with Flow Rate Interaction


The fracture flow rate for the i th discrete fracture given in Eq. A.23 is
qi

V fi
t

Vli N V ji

t
j 1 t

(5)

SPE 140514

DFN Momentum Equations


The pressure loss in terms of the Darcy friction factor based on the cross-sectional average flow velocity from Eq. A.14 is
2
(6)
dp
f v

d
2 dh
Laminar Flow:

The DFN momentum equation (equation of motion) for steady incompressible laminar flow in a narrow ellipsoidal slot
( a b ) with major and minor radii of a and b for a power law fluid from Eq. A.6 is:
n
n
dp
2n 1 k (q a)
(7)

n 2 n1
dx
4n (n) b
The governing fluid front relationship in terms of the slot width ( 2b ) and the pressure differential (pressure loss, p )
from Eq. A.11 for each discrete fracture is
4n (n) t 11 n 1 n
(8)
L L1 n

p
1 n b
2n 1 k
Turbulent Flow:

The pressure loss in terms of the Darcy friction factor for turbulent flow from Eq. A.14 for an elliptical slot is
dp
f q2

dx
2 3 a 2 b3

(9)

where the cross sectional average velocity, v q ab and the hydraulic diameter for a narrow elliptical slot is dh b .
The fluid front propagation relationship for each discrete fracture from Eq. A.16 is
12

2 3

bp t
L L
f

(10)

12

Width-Opening Pressure Constitutive Relationship with Mechanical Interaction


The aperture relationship that includes mechanical interaction at the fracture mid-height (or length) from Eq. A.28 is:
w w

2 H ( p )
E

(11)

where is the fracture induced stress field along the -axis and p p is the net fracture pressure.
The aperture ratio of a uniformly pressurized fracture with mechanical interference in a saturated fracture network from
Eq. A.29 is
( )

w
w (0)

1 p 1

(12)

where the 3D influence function from Eq. A.26 is


h 2
1 1

2d i

3/ 2

(13)

Interior and Exterior Aperture Ratios


The interior and exterior aperture ratios for a finite number of interacting equally spaced fractures are approximated by Eqs.
A.30 and A.31
Interior
i

w (0)

1 p 1

(14)

Exterior
e

w (0)

(1 1 p )

12 (1 )

(15)

SPE 140514

Stiffness Multiplier
The stiffness multiplier (Eq. A.32) is defined as

(16)

where E is an equivalent modulus to correct for mechanical interference (i.e., E E ).


The stiffness multipliers for the interior and exterior fractures are given by Eqs. A.34 and A.35
Interior
i

1
1

(17)

1
1 2 1 2 (1 )

(18)

Exterior
e

Average Dimensionless Width Ratio ( ) and Stiffness Multiplier ( )


The average dimensionless width ratio ( ) and stiffness multiplier ( ) for N fractures ( N 1 ) are given by Eqs. A.36 and
A.37
2 ( N 2)i
(19)
e
N

and

N e i
2 i ( N 2) e

(20)

where for an infinite number of interacting fractures i and i .


Parametric Studies
A few parametric studies will be presented to illustrate the effect of various parameters on the DFN characteristics. The
studies will focus on parametric sensitivities for: 1) mechanical interaction, 2) fracture spacing and number of multiple
fractures, 3) dominant or primary fracture efficiency, and 4) DFN fairway aspect ratio.
Mechanical Interaction
Parametric studies will be presented to investigate the effect of the number of fractures ( N ) and dimensionless fracture
spacing ( d h ) have on the exterior, interior, and average fracture width profiles. Although Eqs. 16-20 were developed for the
interaction at the center of the fracture network, these equations are very good representations of mechanical interaction over
the fracture height and length. As discussed in Appendix A, the characteristic height ( h ) used in the dimensionless spacing
parameter ( d h ) is the smaller of the fracture total height ( H w ) or two wing fracture length ( 2 x f ). See also Warpinski and
Teufel (1987).
Figures 4 and 5 illustrate the effect of the dimensionless fracture spacing ( d h ) on the stiffness multipliers ( ) and
dimensionless width ratios ( ) as a function of the number of equally spaced parallel fractures for the exterior, interior, and
average width characteristics. Figures 4 and 5 demonstrate that as the dimensionless spacing approaches zero ( d h 0 ), the
interior fracture stiffness goes to infinity ( i ) and the interior fracture width ratio goes to zero ( i 0 ), while the
exterior stiffness multiplier approaches a factor of two ( e 2 ) and the width ratio approaches one-half ( e 1 2 ).
Figures 4 and 5 also show that for dimensionless spacings greater than three ( d h 3 ) mechanical interaction is negligible.
The width profiles displayed in Figures 4 and 5 are for a dimensionless spacing of unity ( d h 1 ) with five and three
fractures, respectively. Britt and Smith (2009) suggest that the distance between fractures spaced the same distance as the
created fracture height ( d h 1 ) allows flow resistance to propagate multiple fractures with widths similar to a single
fracture: i.e., widths capable of accepting proppant. Thus when d h 1 , created fracture widths are drastically reduced
relative to that of a single fracture, and these fractures are unlikely to propagate.

SPE 140514

i
h
d

Figure 4 Stiffness multiplier as a function of the number and spacing of equally spaced parallel fractures.

Figure 5 Average dimensionless width ratio as a function of the number and spacing of parallel fractures.
Nolte (1987) presented an analysis of the degree to which the treating pressure (net pressure) would be higher for fully
interacting multiple parallel fractures. Nolte assumed that the total flow ( q ) for N fractures would be the same as the flow for
a single fracture, with the flow rate distributed equally among the N fractures. Nolte also assumed that the total width ( w ) of
N multiple fractures would be approximately equal to the width of a single fracture with the same net pressure. Accounting
for formation elasticity, Nolte found that the net pressure ratio for slot flow of N multiple fully interacting fractures (
d h 0 ) was
pN p1 N 1/ 2

(21)

The corresponding equation for no stiffness interaction is


pN p1 N 1/ 2

(22)

SPE 140514

The general parametric net pressure equation for the PKN 2D model from Meyer (1986) is
p E 2 n 2 q 2 n

1/(2 n 3)

(23)

The net pressure for N multiple interacting fractures based on an average stiffness and equal flow distribution in each
fracture is
1/ (2 n 3)

2 n

2 n 2 q
pN E

(24)

The net pressure ratios for N multiple fractures for full mechanical interaction ( d h 0 , N ) and non-interacting
fractures ( d h , 1 ) are
Fully Interacting
pN
N 2/ (2 n 3)
pN 1

(25)

pN
N 2 n / (2 n 3)
pN 1

(26)

Non-Interacting

Figure 6 shows the net pressure ratio for N multiple fractures as given by Nolte for slot flow from Eqs. 21-22 and Eqs.
25-26 (PKN 2D model) for a Newtonian fluid. Since the exterior fractures will take most of the flow rate for fully interacting
fractures, the maximum net pressure ratio will be much less than indicated by the above analysis. The limiting pressure ratio
for N multiple fully interacting fractures ( d h 0 ) will thus more closely approach the limit for two fractures, since the
interior fractures will have a negligible fracture aperture. Thus the limiting pressure ratio may be better approximated by
pN
22/ (2 n 3)
pN 1

(27)

Figure 6 Net Pressure Ratio for multiple interacting and non-interacting fractures.
Fracture Spacing and Number of Multiple Fractures
This section follows after the work of Cheng (2009) and investigates the effects of fracture spacing ( d or d h ) and the
number of fractures on the fracture exterior (edge) and interior (center and sub-center). Cheng used a discontinuity method to
construct a boundary element model to analyze the stress distributions around multiple transverse fractures and the
geometries of these fractures. The fracture geometry characteristics relevant to this study are the fracture height of 250 feet
and the fracture half-length of 500 feet. A fracture net pressure of 500 psi was assumed and remained constant for all cases.
The fracture characteristics of height and length, and formation properties (Poissons ratio, Youngs modulus, etc.) also
remained constant and resulted in a simulated width at the fracture center of 0.63 inches for a single fracture. Cheng presented

10

SPE 140514

sensitivity studies for the effect of the number of fractures and the effect of the fracture spacing on the fracture geometry
(apertures). The number of multiple transverse (parallel) fractures modeled ranged from one to five with spacings of 50, 100,
and 150 ft. Table 1 summarizes the numerical results of Cheng and the analytical results in this study using Eqs. 14 and 15.
Table 1 - Summary of fracture aperture results as a function of spacing and number.
Fracture Aperture
Cheng (2009)
Meyer & Bazan (SPE 140514)
Edge
Center
Sub-center
Exterior (Edge)
Interior (Center)
e
e
i

width
width
width
width
width
d
N
(d/h)
(ft)
(in)
(in)
(in)
(in)
(in)
n/a
n/a
1
0.63
1.0
1.0
0.63
50

0.2

0.37

0.59

0.05

0.08

0.33

0.53

0.032

0.05

100
100
100
150

0.4
0.4
0.4
0.6

2
3
5
3

0.40
0.40
0.39
0.42

0.63
0.63
0.62
0.67

0.16
0.16
0.25

0.25
0.25
0.40

0.16
-

0.25
-

0.39
0.39
0.39
0.46

0.62
0.62
0.62
0.73

0.15
0.15
0.29

0.24
0.24
0.45

200
250
500
750

0.8
1.0
2.0
3.0

3
3
3
3

0.51
0.54
0.60
0.62

0.80
0.86
0.96
0.98

0.38
0.45
0.58
0.60

0.61
0.72
0.91
0.96

Table 1 illustrates that as the fracture spacing increases, mechanical interference decreases and both the exterior and
interior fractures apertures increase. The effect of fracture spacing is much greater on the interior fractures than on the exterior
fractures especially for small dimensionless spacings. This result is intuitive since the exterior fractures are effectively only
affected by adjacent fractures on the inside. Cheng also stated that the widths of the center fracture and sub-center fractures
are reduced significantly compared to that of the edge fractures. and that Although the center fracture still looks
symmetrical and elliptic-like, the geometry of the edge fractures becomes asymmetrical in the y-direction and is no longer
elliptic-like. This is the result of mechanical interactions among fractures. The stress concentration between the center and
edge fractures suppresses the edge fractures to dilate toward the center fracture. On the other hand, the edge fractures remain
widely open to other sides. Cheng goes on to state that: Although the width of the center fractures is quite similar to that of
the sub-center fracture, the center fracture is not featured with the smallest width... Again this is because the dilation of the
sub-center fracture is suppressed by both the edge and the central fracture. However, the width profile plot for the edge, subcenter, and center fractures in Chengs paper appears to indicate that the center fracture may have a slightly smaller width
profile for the 100 ft spacing case with five parallel fractures.
Eqs. 14 and 15 are assumed to apply over the entire fracture height, although the 3D influence equation, Eq. 13, is strictly
applicable for the width at the center of the fracture. Thus for a uniformly pressured crack, the interior fractures will have an
elliptical width profile while the exterior fractures will be asymmetrical with different elliptical apertures on the exterior and
interior. The asymmetrical behavior of the edge fractures increases as the spacing between parallel fractures decreases.
Figures 7a-12b illustrate the fracture width profiles as a function of fracture spacing and the number of multiple parallel
fractures. For visualization and clarity, the fracture width profiles have been magnified by a factor of 1000. This visualization
magnification factor was chosen to be the same as that used by Cheng to enable direct comparison of numerical results.
Figures 7a, 8a, 12a show the width profiles versus length and Figures 7b, 8b, 12b show the width profiles as a function of
height. These figures demonstrate the impact fracture spacing has on the aperture.
Figures 7a-12b illustrate that for multiple parallel fractures the fracture spacing controls the magnitude of the fracture
width for a given net fracture pressure. As the spacing decreases, the interior fracture widths decrease and become very small
for dimensionless spacing less than 0.2 ( d h 0.2 ). As the number of fractures increases the average fracture width for N
fractures will decrease. The number of parallel fractures, however, has little effect on the interior and exterior fracture
apertures. Thus the representation of three multiple parallel fractures provides the user with a good understanding of the
fracture aperture for the interior and exterior fractures as a function of spacing. It is also illustrated that since the exterior
fractures are only affected by the dilation of the interior fractures, fracture spacing does not affect the edge or exterior fractures
as much as the interior fractures.

SPE 140514

11

d
(d/h=0.4)
Figure 7b Width profiles vs. height.

Figure 7a Width profiles vs. length (d = 100 ft, d/h = 0.4, N = 2).

d
(d/h = 0.4)

Figure 8b Width profiles vs. height.

Figure 8a Width profiles vs. length (d = 100 ft, d/h = 0.4, N = 3).

d
Figure 9a Width profiles vs. length (d = 100 ft, d/h = 0.4, N = 5).

Figure 9b Width profiles vs. height (d/h = 0.4).

12

SPE 140514

Figure 10a Width profiles vs. length (d = 50 ft, d/h = 0.2, N = 3).

Figure 10b Width profiles vs. height (d/h = 0.2).

Figure 11a Width profiles vs. length (d = 150 ft, d/h = 0.6, N = 3).

Figure 11b Width profiles vs. height (d/h = 0.6).

Figure 12a Width profiles vs. length (d = 250 ft, d/h = 1.0, N = 3).

Figure 12b Width profiles vs. height (d/h = 1.0).

SPE 140514

13

During multi-stage cluster treatments, fracture spacing may be a key design consideration to ensure adequate propagation
of all fractures in each cluster, proppant placement in each cluster, and overall productivity from the stage/cluster fractures.
See also Meyer et al. 2010.
Dominant or Primary Fracture Efficiency
The dominant or primary effective fracture efficiency is strongly affected by proppant transport into the secondary fractures.
If all the proppant remains in the dominant fracture, then the secondary fractures behave as sources of fluid loss from the
dominant fracture. The effective fracture efficiency ( ) if all the proppant remains in the dominant fracture from Eq. A.45 is

DFNVpf VDFN

(28)

where DFN is the discrete fracture network efficiency (i.e., DFN VDFN Vt ), V pf is the primary or dominant fracture volume,

VDFN is the DFN volume, and Vt is the total slurry volume injected.
The approximate fraction of pad ( f PAD ) required to prevent a screen-out (e.g., see Nolte (1988) or Meyer & Associates
(2009, 2010)) is generally within the limits given by

(1 )2 f PAD

1
1

(29)

Thus, as the fracture efficiency decreases, the fraction of pad must be increased to prevent a screen-out.
Figures 13a and 13b demonstrate the effect secondary or complex fractures have on the effective efficiency of the primary
fracture. If all the proppant is suspended and remains in the primary fracture (i.e., if no proppant enters into the secondary
fractures) the probability for a screen-out (in the far-field) is high when large DFNs and SRVs are created. And if all the
proppant settles in the primary fracture, then there is a very high probability of packing off and/or screening-out the fracture in
the near wellbore region. Since these two scenarios are not common in many shale formations, some proppant must be placed
in the secondary fracture network. The other possibility is that the hydraulically induced fracture created is a bi-wing fracture
with little proppant settling and no effect of the proppant on propagation.

Figure 13a Dominant fracture efficiency ratio ( DFN ) vs.

Figure 13b Dominant fracture efficiency ratio ( DFN ) vs.

primary fracture ratio ( Vpf VDFN ).

DFN fracture ratio ( VDFN Vpf ).

DFN Fairway Aspect Ratio,


Hydraulically induced fractures generating DFNs and fracture complexity in naturally fractured shale formations have been
shown to create fracture lengths that are much less than that of bi-wing fractures as evidenced by microseismic mapping. This
study will investigate the slurry volume requirement to create a given dominant fracture half-length ( x f a 2 ) of 1000 feet as
a function of the SRV aspect ratio ( b a ) where a and b are the major and minor ellipse axes of the complex fracture
network. Normally, wider fairways should be expected when the minimum and maximum horizontal stresses are close. As
the difference in the maximum and minimum horizontal stresses increases, the fairway will become narrower and the DFN
will approach a single bi-wing fracture.
The following parametric study is based on a set of user specified options. Although the numerical results will differ
depending on the specific options selected, we have set up the DFN numerical simulation as follows: 1) modeling the DFN as
a user specified aspect ratio (versus input of the principal stress differences), 2) assume a saturated or infinite extent of the
DFN system, and 3) the numerical solution will be based on continuum theory (versus discontinuous). The simulations were
run for options of empirical mechanical interaction and no mechanical interaction. Fluid loss interaction was not modeled.
The simulations are based on a slickwater fracturing fluid with an injection flow rate of 100 bpm with no proppant in a
hypothetical shale formation. The numerical results will also illustrate the effect of mechanical fracture interaction on the
DFN characteristics. The fracture spacing in the primary x-z and secondary y-z planes is taken to be equal to the upper and
lower layer thicknesses of about 75 feet (i.e., x y 75 ft ). Horizontal fractures in the x-y plane were not modeled
discretely (i.e., z , with an aperture of zero).

14

SPE 140514

Figure 14 shows the stress and width contour profiles for the dominant fracture with an aspect ratio of one-half ( 0.5 ).
Figures 15 and 16 show 2D and 3D aerial views of the discrete fracture network width profiles (x-y plane), respectively.
Figure 17 shows a 3D major axis view (x-z plane) for the DFN.

Figure 14 Stress, width, and contour profiles for the dominant fracture - b a 0.5 .

Figure 15 DFN 2D aerial view (x-y plane), b a 0.5 .

Figure 16 DFN 3D aerial view (x-y plane).

Figure 17 Discrete Fracture Network (DFN) 3D view (x-z plane), b a 0.5 .


Table 2 shows the volume required to generate a DFN with a major axis of 2000 feet (fracture half-length of 1000 feet) for
various fairway aspect ratios. The simulation results in Table 2 include the effect of mechanical interaction. Table 2 also
presents the results of the numerically generated fracture characteristics. Table 3 shows a corresponding set of results without
mechanical fracture interaction for various fairway aspect ratios.

SPE 140514

15

Tables 2 and 3 illustrate that as the DFN aspect ratio increases, the slurry volume to create a fracture half-length of 1000
feet (major axis of 2000 feet) also increases (see also Figure 17). The volume required to create a bi-wing fracture ( 0 )
with a fracture half-length of 1000 feet is about 60,000 gallons, whereas the volume required to create the same fracture halflength for a very wide DFN fairway ( 1 ) is about 1,200,000 gallons. Tables 2 and 3 show that the maximum fracture width
(at the wellbore) and fracture efficiency decrease (because of increased leakoff area) as the DFN aspect ratio increases. The
general trend of requiring a larger injected volume to create a given fracture length for increasing fracture complexity and
DFN aspect ratio (i.e., increasing DFN area and SRV) is illustrated and expected.
Tables 2 and 3 further demonstrate that when mechanical fracture interaction is taken into consideration fracture width
decreases and fracture height growth (propagation) increases in the upper and lower barriers as a result of an elevated fracture
pressure (i.e., greater net pressure). The created fracture area and total SRV are also greater when mechanical interaction is
considered as a result of the increased fracture height growth. The extent of the increased height growth due to mechanical
interaction of course depends on the magnitude of the confining stress contrasts. If the formation has high upper and lower
confining stress barriers, the volume to create a given DFN fairway would be much less and the net pressure much higher
when mechanical interaction is considered.
Table 2 - Summary of DFN characteristics versus aspect ratio With Mechanical Interaction.
Dominant & DFN
Characteristics
Time, min.
Volume, 1000 gals
Net Pressure, psi
BHFP, psi
Length, ft
Max. Width, in.
Wellbore Height, ft
Efficiency
DFN Area, 106ft2
DFN SRV, 106ft3

0.0
13.9
58
202
6117
1000
0.332
249
0.93
0.4
0.01

Discrete Fracture Network Aspect Ratio, b / a


0.05
0.25
0.5
0.75
29.0
79.9
143.9
213.0
122
335
604
895
200
232
240
236
6115
6147
6155
6151
1000
1000
1000
1000
0.237
0.144
0.114
0.100
247
273
282
280
0.88
0.74
0.63
0.55
1.33
5.4
10.2
15.7
33.5
185.3
382.2
569.6

Table 3 - Summary of DFN characteristics versus aspect ratio Without Mechanical Interaction.
Dominant & DFN
Discrete Fracture Network Aspect Ratio, b / a
Characteristics
0.0
0.05
0.25
0.5
0.75
Time, min.
13.9
29.3
82.7
150.5
223.8
Volume, 1000 gals
58
123
347
632
940
Net Pressure, psi
202
161
117
100
91
BHFP, psi
6117
6075
6031
6014
6005
Length, ft
1000
1000
1000
1000
1000
Max. Width, in.
0.332
0.260
0.190
0.163
0.148
Wellbore Height, ft
249
219
193
185
182
Efficiency
0.93
0.88
0.77
0.67
0.60
DFN Area, 106ft2
0.4
1.18
3.9
7.1
10.3
DFN SRV, 106ft3
0.01
29.8
131.6
253.1
372.6

1.0
289.0
1214
230
6145
1000
0.0924
275
0.49
20.4
746.6

1.0
303.7
1276
85
5999
1000
0.139
180
0.55
13.4
491.2

16

SPE 140514

b/a

Figure 17 Injected volume required to create a DFN major axis half-length of 1000 feet for various DFN aspect ratios.
Case Study Marcellus Shale
Although most conventional fracture treatments result in bi-wing fractures, there are naturally fractured formations that
provide the geomechanical conditions that enable hydraulically induced discrete fractures to be initiated and to propagate in
multiple planes as indicated by microseismic mapping. In these cases, planar hydraulic fracture models may not be the proper
tool to predict fracture propagation.
DFN Numerical Simulation and Pressure Match
A DFN numerical simulation and pressure match for a Marcellus shale well (Well A) is presented. The Marcellus shale is part
of the Devonian black shale group. This case study was published by Jacot et al. (2010) in a paper discussing technology
integration as a methodology to enhance production and economics in horizontal Marcellus shale wells. The paper contains a
detailed description of the petrophysical data, minifrac analysis, DFN model input data, microseismic analysis, and production
results. The reader is, therefore, referred to Jacot (2010) for a detailed explanation of this case study. A brief discussion of the
fracture treatment will be presented here for completeness.
The horizontal Marcellus shale Well A with a lateral length of 2,100 feet was completed in seven fracture stages with five
perforation clusters per stage. Each stage was isolated with a composite bridge plug. The treatment design consisted of
450,000 gallons of slickwater (water, friction reducer and surfactant) and 300,000 pounds of 100 mesh sand and 100,000 lbm
of 40/70 resin coated sand per stage. The treatment design rate was 100 bpm, but higher than expected surface treating
pressures limited the rate to an average of about 85 bpm.
The interrelationship of fracture geometry and areal extent results was evaluated based on flow regimes, wall roughness,
and proppant settling options. Mechanical rock property input values for the DFN simulator were generated from the
geomechanical analysis. The input values include estimates for closure stress, Youngs modulus, Poissons ratio, and fracture
toughness to define the rock properties. The average log-derived closure stress gradient in the lower Marcellus and upper
Marcellus is 0.86 psi/ft and 0.87 psi/ft, respectively.
Although there are no definitive or unique solutions, two very different Discrete Fracture Network simulations were
presented by Jacot. The first DFN simulation scenario assumed each cluster creates a separate network that interferes with the
other clusters. The second scenario assumed that the five clusters in each stage will interact and that they propagate within the
same secondary fracture network. That is, one large fracture network is created independent of which cluster takes most of the
fluid.
These numerical simulations also accounted for extended wellbore storage (EWS) and pressure losses in the fracture midfield resulting from the initiation and fracture complexity around the horizontal wellbore. The fracture initiates as a horizontal
fracture and through the extended near wellbore region, twists and turns, and finally re-orients itself vertically in the principal
planes (see Weng (1993)). This extended wellbore phenomenon (which extends tens of feet) is formulated and presented in

SPE 140514

17

Appendix B and may be the reason why large fracture net pressures and ISIPs are observed with stress gradients of 1.0 psi/ft
or greater. The resulting extended wellbore excess pressure for these simulations was about 3,350 psi, which dissipates in
minutes or tens of minutes after shut-in. The maximum and minimum stress gradients were about 1.0 psi/ft and 0.85 psi/ft.
Therefore, the resulting stress difference at a TVD of 7,800 ft is about 1,200 psi. It is not uncommon to have an ISIP gradient
of 1.3 psi/ft or higher which indicates extended wellbore storage and pressure losses of near 3,500 psi.
The discrete fracture network spacing was assumed to be approximately proportional to the zone thickness. Although the
zones vary, a nominal value of 75 ft was used for both the minor and major axis. Horizontal fractures were not modeled
directly in the simulations but rather accounted for in the extended wellbore pressure loss/storage function as discussed above.
The DFN connected cluster simulation and summary of microseismic mapping results are presented below.
Connected Cluster DFN Simulation and Pressure Match
The connected-cluster simulation assumes that the multi-clusters create a single DFN that interacts with the other clusters.
That is, fractures from the separate clusters may coalesce with other secondary fractures created from the individual clusters.
This assumption may be the best limiting case in that secondary fractures created by individual clusters can occupy the same
secondary network system rather than assuming separate interacting fractures in the principal planes.
Figure 18 illustrates the surface and bottomhole treatment pressure match. The fracture net pressure and dominant fracture
characteristics (width contours, height, length, etc.) are shown in Figures 19 and 20, respectively. Figures 21 through 23 show
various 2D and 3D views of the generated DFN.

Figure 18 Pressure history match for Well A


(Jacot 2010).

Figure 19 Net pressure for Well A (Jacot 2010).

Figure 20 Stress, width, and length profiles for Stage 2 - dominant fracture.

18

Figure 21 DFN 2D aerial view (x-y plane).

SPE 140514

Figure 22 DFN 3D aerial view (x-y plane).

Figure 23 DFN simulation: 3D major axis partial view (x-z plane).


Microseismic Mapping Analysis Marcellus Well
The Well A completion consists of seven fracture stimulation stages. The object of the project was to measure height, length,
and azimuth of the different stimulations versus time. This was done by detecting and locating microseismic events. These
events represent double-couple events (slippage in the rock) that were located based on P-waves and S-waves through the use
of a velocity model calibrated to the perforation shots. A complete description of all seven stages is provided by Jacot (2010).
From a total picture there appears to be activity from the toe to the heel with good extension from the wellbore on both
sides. The far side events are limited to about 2,700 feet growth viewing distance from the monitor well. The actual growth
over time, however, shows trends that are different from the later stages, with the activity occurring with trends parallel to the
wellbore and away from the wellbore. Note that the monitor well placement will favor the near side of the wellbore but should
provide symmetric viewing of microseismic events to the lateral length of the wellbore. Figures 24 and 25 illustrate the side
and top view of microseismic events for Stage 2 and all seven stages, respectively.

SPE 140514

19

Figure 24 Stage 2: Side and top view of microseismic events.

Figure 25 Side and top view of all (seven stages) microseismic events.
Summary of DFN Numerical Simulation and Microseismic Results Marcellus Shale
A summary of the connected cluster DFN characteristics as presented above are given in Table 4 (see Jacot (2010)). The
microseismic mapping analysis is shown for comparison. The connected DFN resulted in greater height growth than the
multi-cluster simulations because of secondary fracture connectivity as reported by Jacot. The connected DFN also matched
better with the microseismic mapping. The pressure history match is essentially identical for the multi-cluster and connected
cluster numerical solutions as given by Jacot (2010). Table 4 illustrates that, as a first order, the DFN characteristics ar e
comparable to the microseismic events. Clearly the DFN has a wide fairway rather than bi-wing behavior, implying the
anisotropic behavior between SHmax( 2 ) and Shmin( 3 ).
Table 4 - Summary of DFN numerical simulation and microseismic mapping Marcellus well.
Simulation

Connected DFN Clusters


Microseismic

DFN Length
(major axis)
(ft)
2200
1800

DFN Width
(minor axis)
(ft)
1100
1200

Fracture
Height
(ft)
340
400

Conclusions
A solution methodology and mathematical formulation for an induced hydraulic Discrete Fracture Network (DFN) numerical
simulator has been presented. The discrete fracture network is composed of discrete and secondary fractures in the three
principal planes. The dominant or primary fractures propagate in the x-z plane perpendicular to the minimum horizontal
stress, . The y-z and x-y plane fractures propagate perpendicular to the and , stresses, respectively. The discrete
fractures created in the x-z and y-z planes are vertical, while the induced fractures created in the x-y plane are horizontal. The
spacing in the x-z, y-z, and x-y planes are y , x , and z , respectively. The formulation utilizes a pseudo-three-dimensional
ellipsoidal approach and includes fracture interaction (fluid loss and mechanical), proppant transport, and planar fracture
propagation in the principal planes. Numerous parametric investigations and a Marcellus shale case study were presented.

20

SPE 140514

A few of the major conclusions and contributions are summarized below:


1. A discrete fracture network model has been formulated. The utility of this model is most applicable in naturally
fractured formations that when hydraulically fractured induce Mode I discrete fracture networks or complex fractures.
2. Mechanical interaction of multiple parallel fractures will have a major impact on the fracture aperture for closely
spaced parallel or transverse fractures. Mechanical interaction has a more dramatic effect on the interior than the
edge or exterior fractures.
3. A new extended wellbore storage and dissipation function was introduced that accounts for the complex fracture
behavior in the extended wellbore or mid-field region of the induced hydraulic fracture network. This extended
wellbore storage function accounts for the high treating pressures and steep pressure decline rates that have been
shown to occur in many horizontal wells and naturally fractured formations that are hydraulically fractured.
4. The fracture efficiency of the dominant fracture has major implications for proppant transport and placement. If all
the proppant is assumed to remain within the dominant fracture, the effective fracture efficiency could be extremely
small for a large DFN or SRV and result in screen-outs. Since screen-outs and proppant packing in the far-field are
not routinely observed it may imply that significant proppant volumes are being transported into the secondary
fracture network.
5. Microseismic mapping can be another useful diagnostic tool in calibrating the DFN fracture model.
6. Application of this technology will provide the operator with a systematic approach for designing, analyzing, and
optimizing multi-stage/multi-cluster transverse DFN induced hydraulic fractures in horizontal wellbores.
Nomenclature
a
= Ellipse major axis
= Area
A
= Ellipse minor axis
b
= Leakoff coefficient
C
= Dimensionless fracture conductivity
C fD
= Hydraulic diameter
dh
= Spacing between discrete fractures
d ij

h
H

=
=
=
=
=
=

Hp

= Pay zone height

Hw
k
L
n
N
p
P

=
=
=
=
=
=
=
=
=
=
=

E
E
f
G( )

pi
q
r
Sp

Young's modulus
Effective Young's modulus
Darcy friction factor
Nolte G function, Fluid loss function
Fracture height
Characteristic fracture half-dimension
Total wellbore height
Consistency index
Fracture half-length
Flow behavior index
Total number of transverse fractures
Pressure
G-function pressure, ISIP p(t )
Reservoir pressure
Injection or flow rate
Radial coordinate
Spurt loss coefficient

= Time
= Closure time, tc t p

tp
v
Vf

= Pump time

Vl
Vsp
w

= Fluid loss volume (no spurt loss)


= Volume loss by spurt

= Velocity
= Fracture volume

= Fracture width

SPE 140514

21

Greek

a
L
p

= Leakoff area parameter


= Length propagation parameter
= Pressure parameter

c
p

= Width propagation parameter

x
y
z

c
,

ij

Subscripts
D
DFN
f
l
p
pf

= Propagation parameter
= G function parameter
= Net fracture pressure, p f
= Discrete fracture network spacing, y-z plane
= Discrete fracture network spacing, x-z plane
= Discrete fracture network spacing, x-y plane
= Fracture efficiency
= Efficiency excluding spurt loss
= Reservoir aspect ratio
= Dimensionless time, t t p
= Dimensionless closure time, c tc t p
=
=
=
=
=

Integration indices
Dimensionless coordinate, x L
Time of fracture leakoff area creation
Fluid loss parameter
3D influence factor

= Fluid loss parameter


= Dimensionless momentum parameter, Stiffness multiplier
= Minimum horizontal stress
=
=
=
=
=
=

Dimensionless
Discrete fracture network
Fracture
Fluid loss
Pay zone, end of pumping
Primary or dominant fracture

Acknowledgments
The authors wish to thank the personnel at Meyer & Associates, Inc. for their support during the preparation of this
manuscript.

References
Adda-Bedia, M. and Amar, Ben M.: Fracture Spacing in Layered Materials, The American Physical Society, Vol. 86, No.
25, June 2001.
Britt, L.K. and Smith, M. B.: Horizontal Well Completion, Stimulation Optimization, and Risk Mitigation, SPE 125526,
September 2009.
Bazan, L.W., Larkin, S.D., Lattibeaudiere, M.G., and Palisch, T.T.: Improving Production in the Eagle Ford Shale with
Fracture Modeling, Increased Conductivity and Optimized Stage and Cluster Spacing Along the Horizontal Wellbore,
SPE 138425, October 2010.
Castillo, J.L.: Modified Fracture Pressure Decline Analysis Including Pressure-Dependent Leakoff, SPE 16417, May 1987.
Cheng, Y.: Boundary Element Analysis of the Stress Distribution around Multiple Fractures: Implications for the Spacing of
Perforation Clusters of Hydraulically fractured Horizontal Wells, SPE 125769, September 2009.
Cipolla, C.L., Warpinski, N.R., Mayerhofer, M.J., Lolon, E.P., and Vincent, M.C.: Relationship between Fracture
Complexity, Reservoir Properties, and Fracture Treatment Design, SPE 115769, September 2008.
Dahi-Taleghani, A. and Olson, J.E.: Numerical Modeling of Multi-Stranded Hydraulic Fracture Propagation: Accounting for
the Interaction between Induced and Natural Fractures, October 2009.

22

SPE 140514

Davidson, B.M., Saunders, B.F., Robinson, B.M., and Holditch, S.A.: Analysis of Abnormally High Fracture Treating
Pressures caused by Complex Fracture Growth, SPE 26154, June 1993.
Engelder, T. and Lash, G.: http://geology.com/articles/marcellus-shale.shtml, July 7, 2008.
Engelder, T., Lash, G.G., Uzcategui, R.S.: Joints sets that enhance production from Middle and Upper Devonian gas shales of
the Appalachian Basin, AAPG Bulletin, V. 93, No. 7 pp. 857-889, July 2009.
Fontaine, J., Johnson, N., and Schoen, D.: Design, Execution, and Evaluation of a Typical Marcellus Shale Slickwater
Stimulation: A Case History, SPE 117772, October 2008.
Hainey, B.W. and Weng, X.: Mitigation of Multiple Fractures from Deviated Wellbores, Oct. 1995.
Jacot, R.H., Bazan, L.W., and Meyer, B.R.: Technology Integration A Methodology to Enhance Production and Maximize
Economics in Horizontal Marcellus Shale Wells, SPE 135262, September 2010.
King, G.E., Haile, L., Shuss, J., and Dobkins, T.: Increasing Fracture Path Complexity and Controlling Downward Fracture
Growth in the Barnett Shale, SPE 119896, November 2008.
Mayerhofer, M.J., Lolon, E.P., Warpinski, N.R., Cipolla, C.L., and Walser, D.: What is Stimulated Rock Volume? SPE
119890, November, 2008.
Meyer & Associates, Inc.: Meyer Fracturing Simulators User's Guide, Seventh, 2009, and Eighth Edition, 2010.
Meyer, B. R.: Design Formulae for 2-D and 3-D Vertical Hydraulic Fractures: Model Comparison and Parametric Studies,
SPE 15240 presented at the SPE Unconventional Gas Technology Symposium, Louisville, KY, May 18-21, 1986.
Meyer, B. R.: Three-Dimensional Hydraulic Fracturing Simulation on Personal Computers: Theory and Comparison
Studies, paper SPE 19329 presented at the SPE Eastern Regional Meeting, Morgantown, Oct. 24-27, 1989.
Meyer, B.R. and Hagel, M.W.: "Simulated Mini-frac Analysis," JCPT (Sept./Oct. 1989), Vol. 28, No. 5, 63-73.
Meyer, B.R. and Jacot, R.H.: Implementation of Fracture Calibration Equations for Pressure Dependent Leakoff, SPE
62545, June 2000.
Meyer, B.R., Bazan, L.W., Jacot, R.H., and Lattibeaudiere, M.G.: Optimization of Multiple Transverse Hydraulic Fractures
in Horizontal Wellbores, SPE 131732, February 2010.
Muthukumarappan, R., Lyons, B., Hendrickson, R.B., Barree, R.D., and Magill, D.P.: Effects of High-Pressure-Dependent
Leakoff and High-Process-Zone Stress in Coal Stimulation Treatments, SPE 107971, April 2007.
Muthukumarappan, R., Barree, R.D., Broacha, E., Barett, B., Longwell, J.D., Kundert, D.P, and Tamayo, C.: Effects of HighProcess-Zone Stress in Shale Stimulation Treatments, SPE 123581, April 2009.
Nolte, K.G.: Determination of Fracture Parameters from Fracture Pressure Decline, SPE 8341 presented at the 54th Annual
Technical Conference, Las Vegas, Sept., 1979.
Nolte, K.G.: A General Analysis of Fracturing Pressure Decline with Application to Three Models, (SPE 12941) JPT (Dec.
1986), 571-582.
Nolte, K.G.: Discussion of Influence of Geologic Discontinuities on Hydraulic Fracture Propagation, (SPE 17011) JPT
(Aug. 1987), 998.
Nolte, K.G.: Fracture Design Considerations Based on Pressure Analysis, SPEPE (Feb. 1988), 22-30.
Nolte, K.G.: Application of Fracture Design Based on Pressure Analysis, SPEPE (Feb. 1988), 31-41.
Nolte, K.G.: Fracture Pressure Analysis for Non-Ideal Behavior, (SPE 20704) JPT (Feb. 1991), 210-218.
Nolte, K.G., Mack, M.G. and Lie, W.L.: A Systematic Method of Applying Fracture Pressure Decline: Part I, SPE 25845
presented at the SPE Rocky Mountain Regional/Low Permeability Symposium held in Denver, CO, April 12-14, 1993.
Olson, J.E.: Fracturing from Highly deviated and Horizontal Wells: Numerical Analysis of Non-planar Fracture
Propagation, SPE 29573, March 1995.
Olson, J.E.: Sublinear scaling of fracture aperture versus length: An exception to the rule?, J. Geophysical Research, Vol.
108, No. B9, p. 3-1 to 3-11, 2003.
Olson, J.E.: Predicting fracture swarms the influence of subcritical crack growth and the crack-tip process zone on joint
spacing in rock, Geological Society, London, Special Publications, 2004; v. 231; p. 73-88.
Olson, J.E.: Spatial Organization of Natural Fractures: A Geomechanics Approach, http://www.pe.utexas.edu/~jolson/utigtalk_files/v3_document.htm & Natural Fracture Pattern Development, http://www.pe.utexas.edu/~jolson/nat.frac.html.,
2008-2010.
Peidro, F.: The Relationship between Fracture Density and the Mechanical Layering of Rocks, Norwegian University of
Science and Technology, 2009.
Pollard, D.D. and Segall, P.: Theoretical displacements and stresses near fractures in rock: With applications to faults, joints,
veins, dikes and solution surfaces, Fracture Mechanics of Rocks, Academic, pp. 277-350, 1987.

SPE 140514

23

Sagy, A. and Reches, Z.: Joint intensity in layered rocks: The unsaturated, saturated, supersaturated, and clustered classes,
Institute of Earth Sciences, Hebrew University, 2004.
Tang, C.A., Liang, Z.Z., Zhang, Y.B., Chang, X., Tao, X., Wang, D.G., Zhang, J.X., Liu, J.S., Zhu, W.C., and Elsworth, D.:
Fracture Spacing in Layered Materials: A New Explanation based on Two-dimensional Failure Process Modeling,
American J. of Science, Vol. 308, p. 49-72, January 2008.
Warpinski, N.R. and Teufel, L.W.: Influence of Geologic Discontinuities on Hydraulic Fracture Propagation, (SPE 13224)
JPT 209-220, Feb. 1987.
Warpinski, N.R., Mayerhofer, M.J., Vincent, M.C., Cipolla, C.L., and Lolon, E.P.: Stimulating Unconventional Reservoirs:
Maximizing Network Growth while Optimizing Fracture Conductivity, SPE 114173, February 2008.
Warren, J.E. and Root, P.J.: The Behavior of Naturally Fractured Reservoirs, Soc. Pet. Eng. J., (SPE 426), Sept. 1963, pp.
245-255.
Weijers, L.G., Sugiyama, H., Shimamoto, T., Takada, S., Chong, J.M., and Wright, C.A.: The First Successful Fracture
Treatment Campaign Conducted in Japan: Stimulation Challenges in a Deep, Naturally Fractured Volcanic Rock, SPE
77678, Oct. 2002.
Weng, X.: Fracture Initiation and Propagation from Deviated Wellbores, SPE 26597, Oct. 1993.
Xu, W., Thiercelin, M., and Walton, I: Characterization of Hydraulically-Induced Shale Fracture Network Using an
Analytical/Semi-Analytical Model, SPE 124697, Oct. 2009.
Yeager, B.B. and Meyer, B.R.: Injection/Fall-off Testing in the Marcellus Shale: Using Reservoir Knowledge to Improve
Operational Efficiency, SPE 139067, Oct. 2010.
Zoback, M.L.: First and second order patterns of stress in the lithosphere, The World Stress Map Project: Journal of
Geophysical Research, v. 97, p. 11, 703-11, 728, 1992.
Appendix A: A Discrete Fracture Network (DFN) Model
The solution methodology for our Discrete Fracture Network (DFN) Simulator (MShale) is formulated in The Meyer User's
Guide (2009) and presented here in detail. The fractures are assumed to be discrete and may or may not interact.
Although most fracturing applications consist of bi-wing fractures because they require less energy to propagate, there are
natural systems that force the boundary condition to create discrete fractures requiring even less energy than a bi-wing. This is
most easily understood by examining limited entry designs or by perforating large intervals in highly deviated wellbores. Both
of these are examples of creating multiple fractures as the result of a boundary condition.
The fundamental first-order DFN mass and momentum conservation equations are based on a self-similar solution
methodology. The formulation utilizes a pseudo-three-dimensional ellipsoidal approach. The major assumptions are:
1. The dominant fracture or main fracture is in the x-z plane and propagates perpendicular to the minimum horizontal stress,
3 . The y-z and x-y plane fractures propagate perpendicular to 2 and 1 , respectively.
2. The discrete fracture network may be composed of secondary fractures in all three principal planes. The spacings in the
x-z, y-z, and x-y planes are y , x , and z , respectively.
3. The boundary conditions are such that the natural fracture system can initiate multiple hydraulic fractures. This is similar
to creating multiple fractures from highly deviated vertical wellbores with large perforated intervals. This forces the
condition to propagate multiple fractures in various planes in preference to a single bi-wing dominant fracture as is typical
for limited type entry designs.
4. Fractures will only propagate in the y-z and x-y planes if the fracture pressure is greater than the corresponding minimum
horizontal stress in that plane.
5. Fractures in the x-z plane (other than the dominant fracture) cannot open unless a fracture network in the y-z plane is
established for the fracture to propagate (i.e., the fractures must be connected in the DFN). These assumptions are not
applicable for multiple or cluster type fractures which may be disjoint.
6. The numerical solution will be based on ellipsoidal self-similar type equations. That is, the fracture stimulated reservoir
volume will be ellipsoidal as will the geometric distributions. The width and height profiles, however, are calculated from
the governing P3D pressure-width-height relationships.
7. Mechanical and fluid loss interaction of multiple parallel fractures is considered.
8. The fracture height at the joints in the x-z, and y-z planes will be assumed to be the same. This assumption is true for 2D
and penny shaped fractures but may not be true for well contained fractures.
9. Dilatancy of interacting fractures at the joints is ignored.
10. The fracture network extension (with the exception of the dominant fracture) can be limited to a finite fracture extent in
each plane.

24

SPE 140514

11. The discrete fractures created in the x-y plane are assumed to be horizontal with the same aspect ratio as the y-z plane
fracture.
The discrete fracture network model is based on a similar methodology presented by Warren and Root (1963) for dual
porosity natural fractured reservoirs. Figure A.1 illustrates the idealization of a heterogeneous porous media and depicts an
independent system of secondary porosity (that) is superimposed on the primary or intergranular system. The DFN model in
this paper accounts for P3D fracture growth in the three principal planes and includes options for proppant transport
distribution throughout the primary and secondary fracture network.

Figure A.1 Warren and Root (1963) - Idealization of the heterogeneous porous media.
The fracture network model assumes primary and secondary fractures can be initiated and propagate discretely in the
principal planes provided the conditions discussed above are satisfied. That is, vertical fractures can be initiated in the x-z
plane (major axis) and y-z plane (minor axis), and horizontal fractures can propagate in the x-y plane (vertical axis). Figures
A.2 through A.4 demonstrate typical fracture profiles generated for an induced DFN in the three principal planes. As
illustrated, the horizontal DFN fractures are in the x-y plane and vertically induced fractures are created in the y-z (minor axis)
and x-z (major axis) planes. The minimum horizontal stress, 3 , is perpendicular to the major axis (x-z plane).

2 3 200 psi

x 100 ft
y 75 ft

w ( x, y, z ) w

H ( p )
E

Figure A.2 Numerical DFN Simulation: 2D Aerial (x-y plane) View.

SPE 140514

Figure A.3 Numerical DFN Simulation: 3D Aerial (x-y plane) Partial View.

Figure A.4 Numerical DFN Simulation: Minor Axis (y-z plane) Partial View.

Figure A.5 Numerical DFN Simulation: Major Axis (x-z plane) Partial View.

25

26

SPE 140514

Governing Equations
The governing mass, momentum, and energy equations and constitutive relationships are presented below.
Mass Conservation
The governing mass conservation equation, for an incompressible slurry in the fracture, states that the volume of slurry
injected into the fracture minus the volume loss by leakoff, Vl (t ) , and spurt loss, Vsp (t ) , must be equal to the fracture volume,
V f (t )

t
0

q( ) d Vl (t ) Vsp (t ) V f (t )

(A.1)

Please refer to Meyer (2009) for a detailed discussion of Eq. A.1.


Continuity
The governing continuity equation, for incompressible slurry in the fracture, in terms of the fluid velocity, v , is
vw 2vl w t 0
where vl is the fluid leak-off velocity and w is the fracture width at any position.
Momentum Equation
The momentum equation (equation of motion) can be written as:
Dv

p g
Dt
Simplifying Eq. A.3 for steady flow we have
1 f v 2
p
2 dh

(A.2)

(A.3)

(A.4)

where f is the Darcy friction factor which is a function of the Reynolds Number, Re , and fracture wall roughness.
Width-Opening Pressure Constitutive Relationship
The crack-opening (aperture) and opening pressure relationship from Meyer (1986, 1989) for fractures in each of the principal
planes ( ) is of the form:
2 H ( p )
2 H p
(A.5)
w ( x, y, z ) w

where w is a generalized influence function, H is a characteristic fracture half-height (or length), w is the fracture width,

p is the fracture pressure, E E 2 1 2 is the effective Youngs modulus, is the confining stress and p p
is the net fracture pressure for each discrete fracture in the three principal planes ( ).
To account for mechanical interaction of parallel fractures, Eq. A.5 must be modified as discussed and presented below.

DFN Equations
The fundamental Discrete Fracture Network (DFN) solution methodology is based on satisfying the continuity, mass
conservation, constitutive relationships and momentum equations. These equations are then solved numerically. To illustrate
the methodology, a set of equations will be presented for flow in an ellipsoidal slot. Although the pseudo-three dimensional
equations are more involved, the following will provide the reader with a general understanding of the principles used in the
DFN solution.
DFN Momentum Equations
Laminar Flow:
The DFN momentum equation (equation of motion) for steady incompressible laminar flow in a narrow ellipsoidal slot
( a b ) with major and minor radii of a and b for a power law fluid from Eq. A.4 can be written as:
n

n
dp
2n 1 k (q a)

n 2 n1
dx
4n (n) b

where
(n) (1 2 ) 1d

1
1 1 2
B 1 2,
2
2 1 2

and 4n 1 2n .
Rearranging Eq. A.6 in terms of the net fracture flow rate, we find

(A.6)

(A.7)

SPE 140514

27

2 1 n
(A.8)
4n (n)ab
p n
q

1 n
k
2n 1
L
where L is the length of the fluid front in an elliptical slot. The net slot flow rate, q , for a constant cross-sectional area, ab ,
is
L
(A.9)
q ab
t
Substituting Eq. A.7 into Eq. A.6, we find
1

(A.10)
L 4n (n)b11 n p n

1 n
t 2n 1 k
L
where L is the change in the fluid front position over the time step t .
The governing fluid front relationship in terms of the slot width ( 2b ) and the pressure differential (pressure loss, p ) is

4n (n) t 11 n 1 n
L L1 n

p
1 n b
2n 1 k
The fluid fronts for two slots with different slot widths and pressure drops from Eq. A.11 is
L2 L12 n L1 L11 n

(A.11)

(A.12)

where
b 1 n p
2 2
b
p1
1

1 n

(A.13)

Turbulent Flow:
The pressure loss in terms of the Darcy friction factor is
dp
f v

dx
2 dh

f q2
2 3 a 2 b3

(A.14)

where the cross sectional average velocity v q ab , and the hydraulic diameter for a narrow elliptical slot is dh b .
The flow rate from Eq. A.14 is
12

2 3 a 2b3 p
q

L
f
Substituting the slot flow rate from Eq. A.15 into Eq. A.14, we find

12

2 3

bp t
L L
f

The fluid fronts with turbulent flow for two slots with different widths and pressure drops from Eq. A.16 is
L2 L12 2 L1 L11 2
where
12

12

b p
2 2
b1 p1

(A.15)

(A.16)

(A.17)
(A.18)

Time Dependent Cross-Sectional Area:


The above equations are easily modified for a time dependent ellipsoidal cross sectional area (and flux). The power law
behavior for pseudo-steady state 2D and 3D fracture propagation is of the form (see Meyer (1986, 1989)):
t d
(A.19)

dt
where is a characteristic fracture dimension (e.g., a, b, L ).
The net flow rate for a time dependent geometry (cross-sectional area and length) is
L
(A.20)
q ab
1 a b
t
where a a L and a b L .

28

SPE 140514

Thus, the above momentum equations for laminar and turbulent flow are still valid for time dependent height, 2a , and
aperture, 2b .
DFN Interaction
The fluid loss and mechanical (stiffness) interactions of discrete fractures are given below.
Fluid loss
Fluid loss interaction accounts for the reduction in fluid leakoff as a result of competing discrete fracture interference. The
DFN fluid loss due to leakoff and spurt loss for N discrete fractures of surface area A from Eq. A.1 is
N

i 1

Vl (t ) 2

A
0

v c da d 2 s p c da
A

(A.21)

i 1

The fluid loss multiplier l is defined as:

l ( A / ADFN 1)l 1
where l is the fluid loss interaction factor. If l 0 there is no interaction, l 1 , and for l 1 , the fractures fully
N

interact, l A / ADFN . The total discrete fracture area is given by ADFN Ai .


i 1

Flow Rate Interaction


The fracture flow rate for the i th discrete fracture is given by
qi

V fi
t

Vli N V ji

t
j 1 t

(A.22)

The total DFN flow rate is


N

qt
i 1

V fi
t

i 1

Vli
t

(A.23)

Mechanical Interaction
Mechanical interaction between fractures occurs when fractures are close enough to be affected by the stress field from
adjacent fractures. The fracture induced stress field xx (see Pollard & Segall (1987), Olson (2004), Warpinski & Teufel
(1987)) along the x-axis at the mid-height or center of the fracture can be written as
xx


h
I 1 1
2d
ij

3/ 2

(A.24)

where I is the Mode I driving stress, h is the total fracture height and dij is the distance between the center of parallel
fractures i and j . The induced fracture stress field is compressive and asymptotes to the driving stress as fractures become
very close. The Mode I driving stress (Olson (2003, 2004), Warpinski & Teufel (1987)) is defined for uniformly loaded cracks
as
I p f

(A.25)

where p f is the fracture pressure and is the remote normal stress (or minimum horizontal stress) perpendicular to the
fracture face. Therefore, the Mode I driving stress is equal to the fracture net pressure (i.e., p I p f ).
The 3D stress influence function (3D influence factor or 3D correction factor) ij by which the plane-strain influence
functions are multiplied (see Olson (2004)) is then defined as
h
xx
ij
1 1
2dij
I

3/ 2

(A.26)

Thus, the 3D influence factor ij represents the influence of fracture height (or mechanical layer thickness) h on the induced
stress at the center of fractures a distance of dij apart.

SPE 140514

29

Figure A.2 illustrates the magnitude of the 3D influence factor (Eq. A.26) as a function of the dimensionless spacing
(distance) between parallel fractures. Figure A.2 demonstrates that if the fractures are very close ( d h 0 ) the fractures fully
interact but as the fracture spacing increases the mechanical influence decreases. For a dimensionless distance greater than
three ( d h 3 ), multiple parallel fractures have negligible interaction.

Figure A.2 Elastic influence interaction factor versus dimensionless distance.


Width-Opening Pressure Constitutive Relationship with Mechanical Interaction
The fracture aperture without mechanical interference as a function of net pressure (Eq. A.5) in each of the principal planes (
) is
2 H ( p )
2 H p
(A.27)
w (0) w

The aperture relationship that includes mechanical interaction at the fracture mid-height (or length) from Eq. A.27 is:
(A.28)
where is the fracture induced stress field along the -axis and p p is the net fracture pressure. The above
equation assumes fracture interference on both fracture faces.
The aperture ratio of a uniformly pressured fracture with mechanical interference in a saturated fracture network is
( )

w
w (0)

1 p 1

(A.29)

Interior and Exterior Aperture Ratios


The interior and exterior aperture ratios for a finite number of interacting equally spaced fractures is approximated by
Interior
w

1 p 1

(A.30)

1/ 2(1 1 p ) 1/ 2 1/ 2(1 )

(A.31)

w (0)

Exterior
e

w (0)

Thus, for fully interacting fractures ( 1 ), i 0 and e 1/ 2 , and for non-interacting fractures ( 0 ),
i 1 and e 1 .

30

SPE 140514

Stiffness Multiplier
The stiffness multiplier is defined as

(A.32)

where E is the equivalent modulus to correct for mechanical interference (i.e., E E ). That is,
( )

w
w (0)

E
1

(A.33)

The stiffness multipliers for the interior and exterior fractures are given by
Interior
i

1
1

(A.34)

1
1 2 1 2 (1 )

(A.35)

Exterior
e

Thus for fully interacting fractures ( 1 ), i and e 2 while for non-interacting fractures ( 0 ),
i 1 and e 1 .
Average Dimensionless Width Ratio ( ) and Stiffness Multiplier ( )
The average dimensionless width ratio ( ) and stiffness multiplier ( ) for N fractures ( N 1 ) are
2 ( N 2)i
(A.36)
e
N

and
N / (2e ( N 2)i )

N e i
2 i ( N 2) e

(A.37)

where for an infinite number of interacting fractures ( N ) i and i .


DFN Characteristics
The DFN solution methodology is based on solving the fundamental mass and momentum conservation equations presented
above. The fracture propagation in the three principal planes is governed by the momentum and mass conservation equations
for Mode I failure (i.e., fracture propagation is based on the condition that the fracture pressure exceeds the confining stress
and fracture toughness criteria in that principal direction). The DFN geometric characteristics are summarized below:
LDFN

L j

j 1

ADFN A j
x , y , z j 1

VDFN V j
x , y , z j 1

wDFN VDFN ADFN


x, y, z

(A.38)

where
L d
A h d
V w dA
w V A

The fracture length and width distribution for each network perpendicular to the direction is given by

(A.39)

SPE 140514

31

LD 1 D2

12

wD 1

(A.40)

2 12
D

where LD L / L and wD w / w . The fracture height profile in the secondary plane is assumed to be self-similar with the
main fracture as given by

hD f D

(A.41)

where hD h h (e.g., for an ellipsoidal profile hD 1 D2 ). Although the assumption of an elliptically shaped stimulated
12

reservoir volume is not necessary for a unique solution, it greatly reduces the numerical complexity when solving the
momentum and mass conservation equations.
The x-y plane fracture is horizontal with an ellipsoidal shape and width profile. This configuration allows for T shaped
DFN geometries.
Proppant Distribution
The distribution of proppant in a DFN is a very complicated and inherently difficult problem to solve as discussed by Meyer &
Associates, Inc. (2009, 2010). The proppant may bridge- or screen-out in one fracture (or in a few secondary fractures) while
the others continue propagating. The flow distribution changes with time, and the proppant concentration distribution depends
on individual fracture fluid loss and secondary fracture extension. To handle this problem correctly a fully 3D DFN simulator
that allows individual fractures in the network to bridge- or screen-out and stop propagating while the others remaining
propagating with the redistributed slurry flux is required. Also, if some fractures take only fluid and no proppant, the proppant
will concentrate at a faster rate in the fractures that take the proppant. The dominant fracture efficiency is used to determine
proppant transport and distribution. The flux is also slightly different in the primary fracture depending on the proppant
distribution option and will therefore give slight deviations in the numerical solution.
To simplify the theory (and the calculations) while preserving the limiting solutions, a Proppant Distribution Style for
Uniform, Dominant, and User Specified proppant allocation will be presented. The limiting solutions for uniform distribution
of proppant in the fracture network and for all the proppant remaining in the main or dominant fracture will be discussed.
The proppant distribution allocation is defined as

p M df M DFN M df M t

(A.42)

where M df is the proppant mass in the dominant or primary fracture and M t the total proppant mass injected into the DFN or
fracture system.
This proppant distribution methodology for determining fluid loss and proppant distribution in the primary or dominant
fracture is fundamentally governed by the discrete fracture fluid efficiencies. That is even with no fluid leakoff from the DFN,
the proppant can screen-out in the primary fracture if proppant does not transport into the secondary. Again the full set of
equations and solution technique provides a methodology for various scenarios of proppant distribution and transport in the
primary and secondary fractures.
The minimum proppant allocation is shown to be equal to the ratio of the dominant or primary fracture to the total DFN
volume. That is:

min

Vpf VDFN

(A.43)

An effective fracture efficiency, , for a fraction of the proppant remaining in the primary or dominant fracture ( p ) is

Vpf

V
p t

DFN

Vpf

V
p

DFN

(A.44)

The dominant fracture efficiency for all proppant remaining in the primary fracture ( p 1 ) from A.44 is

DFNVpf VDFN

(A.45)

Appendix B: Extended Wellbore Dissipation and Storage


Weijers et al. (2002) discussed three regions within the fracture: the near-well, mid-field and far-field, with complex, tortuous
and simple fracture characteristics each having unique pressure signatures. Muthukumarappan et al. (2007, 2009) states
screen-outs during shale and coal hydraulic fracturing treatments can be attributable to either high pressure dependent leakoff
(PDL), high process zone stress (PZ) or both.
A new pressure dissipation and extended wellbore storage (EWS) function is formulated that explains the complex
fracture behavior, high fracture pressure (ISIPs) and pressure fall-off as a result of fracture storage in the extended or midfield region of the fracture. The governing equations are presented below.

32

SPE 140514

The near-wellbore pressure loss due to the viscous dissipation as a result of multiple complex fractures and tortuosity in
the near wellbore region is represented by
pnw k (t )q(t ) (t )

(B.1)

where k (t ) is a time dependent proportionality constant and (t ) is a flow rate power coefficient. The power coefficient as a
result of viscous dissipation is theoretically equal to the flow behavior index (i.e., n ) but is generally represented by
1 2 . The pressure loss or pressure dissipation through the perforations is inertial and therefore proportional to the flow
rate squared

p perf k (t )q(t )2

(B.2)

The momentum equation for frictional pressure dissipation in the far-field from Eq. A.4 is
p

1 f v 2
2 dh

(B.3)

where f is the Darcy friction factor.


It has been observed (e.g., Weijers et al. (2002), Weng (1993), and Jacot et al. (2010)) that complex fracture behavior may
occur when fracturing highly deviated and horizontal wellbores. The turning and twisting of complex fractures in the extended
or mid-field region and then re-orientating in the direction of the principal planes creates a high fracture pressure that does not
diminish instantly when the fracture is shut-in. The fracture gradient during pumping can be much greater than the overburden stress gradient without creating horizontal fractures in the far-field.
Figures B.1 and B.2 illustrate the conceptual complex fracture behavior in the mid-field and the far-field discrete fracture
network (DFN) as shown by Weng (1993) and Weijers et al. (2002). Figure B.2 shows a conceptual picture of mid-field
tortuosity. Weijers et al. states Mid-field tortuosity is recognized by a high apparent ISIP and rapidly declining pressures
during the first few minutes of shut-in. This is caused by a pressure choke beyond the near-wellbore area in the fracture.

Weijers et al. (2002)

Fractures link-up by rotating their leading edges until


they are parallel to the wellbore (Weng (1993)).

Figure B.1 Extended or mid-field wellbore storage


schematic (Weng (1993)).

Figure B.2 Conceptual picture of mid-field tortuosity.

The governing equations of mass conservation, compliance, width-opening pressure and constitutive relationships
governing the complex fracture behavior in the extended wellbore region and fracture closure are presented below. The Nolte
G function analysis is based on the original work of Nolte et al. (1979, 1986, 1987, 1988, 1991, 1993), Castillo (1987), and the
methodology of Meyer and Hagel (1989).

SPE 140514

33

Mass Conservation
The governing mass conservation equation for an incompressible slurry in the fracture states that the volume of slurry injected
minus the volume loss by leakoff, Vl (t ) , and spurt loss, Vsp (t ) , must be equal to the fracture volume, V f (t )

t
0

q( ) d Vl (t ) Vsp (t ) V f (t )

(B.4)

where
Vl (t ) 2

t
0

A
0

v dA dt

Vsp (t ) 2 S p A(t )
v(t )

(B.5)

C ( A, t )
t ( A)

( A) t A / A(t )

1/

v is the leakoff velocity and is equal to for square-root loss.


The fluid loss volumes during pumping and after shut-in are presented below.

Fluid Loss - During Pumping


The volume of fluid loss due to leakoff (excluding spurt loss) from Eq. B.4 is

Vl (t ) 2

t
0

A
0

v dAdt C (t ) ADFN (t ) AEWS (t ) t ( a , c )

(B.6)

where is formulated from incomplete gamma functions as given by Meyer and Hagel (1989). The total leakoff coefficient
C (t ) is the instantaneous time dependent value. For a constant leakoff coefficient independent of time and pressure, c is
equal to zero. The fracture area for the extended wellbore storage contributing to fluid loss is given by AEWS , and the fracture
leakoff area of the discrete fracture network is given by ADFN .
Fluid Loss - After Pumping
After pumping (i.e., during closure) the fracture is assumed to stop propagating. The total fluid loss volume after pumping,
Vl (t ) , from Eq. B.6 is

Vl (t ) 2

t
tp

A
0

v dAdt C (t p ) ADFN (t ) AEWS (t ) t p G ( a , ) 2

(B.7)

Vl (t p )G / (2)
where

G 4 a
1

a
0

1/ 1

1/ a

1
2

d d

(B.8)

Eq. B.8 is a general form of the Nolte G function for a constant fluid loss coefficient and simplifies to Noltes original
equation for discrete values of a . The solution for time dependent fluid loss is presented in the Meyer Users Guide, Meyer
& Associates, Inc. (2009). During closure, square root fluid loss ( =1/2) with a constant leakoff coefficient is assumed for
this derivation.
The dimensionless time, , is defined as

t tp

(B.9)

where t p is the pump time.


Extended Wellbore Storage (EWS) Region
Shut-in Mass Conservation
The mass conservation during shut-in for the extended wellbore fracture system and DFN is now addressed. During
closure it is assumed that fluid loss from the extended wellbore storage region is governed by fluid loss from the EWS area
( AEWS ) and flow into the DFN system. Mass conservation during shut-in requires that the fracture volume during shut-in be
equal to the fracture volume at the end of pumping minus the fluid loss after pumping due to leakoff and fluid flow into the
DFN:

VEWS (t ) VEWS (t p ) V

EWS

(t ) Vq (t )

(B.10)

34

SPE 140514

where the fluid leakoff volume from the EWS and the fluid loss volume to the DFN are given by
V

C (t p ) AEWS (t p ) t p G( a , ) 2

EWS

(B.11)

and
Vq (t p t tc ) q f ( )d C (t p ) ADFN (t p ) t p G( a , ) 2
t

(B.12)

tp

The fracture volume ratio during shut-in in the EWS region as a function of time is found by substituting Eq. B.11 and Eq.
B.12 into Eq. B.13 and assuming a constant leakoff coefficient ( C (t t p ) C (t p ) )
VEWS (t ) VEWS (t p ) C (t p ) ADFN (t p ) AEWS (t p ) t p G( a , ) 2

(B.13)

or

VEWS (t )
(1 s ) VDFN (t p ) VEWS (t p )
1

G( )
VEWS (t p )
2 cs
VEWS (t p )

(B.14)

where c ( a , c ) for a constant fluid loss coefficient and s is the fracture efficiency at the end of pumping excluding
spurt loss. This efficiency is defined as

s (t p ) (t p ) / 1

q d

Vsp ( t p )

tp
0

(B.15)

where (t p ) is the fracture efficiency at the end of pumping, t p , and is given by


(B.16)

The resulting closure condition from Eq. B.14 is


(B.17)
Pressure Decline
Assuming the fracture compliance remains constant during closure (i.e., self-similar closure), we have
pEWS (t ) pEWS (t p )

VEWS (t )
VEWS (t p )

(B.18)

where pEWS (t ) = p f (t ) is the fracture net pressure.


The governing fracture pressure decline equation during closure is obtained by substituting Eq. B.18 into B.14 resulting in

(1 s )
pEWS (t )
VDFN (t p ) VEWS (t p )

G( )
2 cs
V
(
t
)
pEWS (t p )

EWS p

(B.19)

Placing Eq. B.19 in terms of the ISIP (i.e., pEWS (t p ) ISIP ) we find
ISIP p(t ) G( )

dP
dG

(B.20)

where

1 s
dP
pEWS (t p )
dG
2 cs
and P ISIP p(t ) .

VDFN (t p ) VEWS (t p )

VEWS (t p )

(B.21)

SPE 140514

35

Pressure Dependent Fluid Loss and/or Compliance


Assuming fluid loss in the EWS region is pressure dependent or the compliance is pressure dependent Eq. B.21 can be written
as follows (see Meyer et al. 2000):
1 s
dP
pEWS (t p )
dG
2 cs

VDFN (t p ) VEWS (t p )

p(t ) p

VEWS (t p )

p(t p ) p

cp

(B.22)

where p is a characteristic pressure (i.e., p pi for pressure dependent fluid loss) cp 1/ 2 for filtrate controlled fluid
loss, cp 1 for reservoir controlled fluid loss, cp 0.2 for ridged filter cake controlled leakoff, and cp 1/ 2 compressible
filter cake controlled leakoff. Other values for this power coefficient are given by Meyer (2000).
Integrating Eq. B.22 yields a dimensionless pressure function of the form

1 s

2 cs

VDFN (t p ) VEWS (t p )
G( )

VEWS (t p )

(B.23)

where

p(t ) p

1
ISIP p

1 cp

p(t ) EWS
ISIP p

/ 1 cp

(B.24)

(B.25)

Generalized EWS Pressure Decline Function


The complex fracture geometry and pressure behavior (decline) in the EWS region can be a very complicated phenomenon to
mathematically represent for all cases (e.g., pressure dependent compliance, pressure dependent fluid loss and disipation of
energy in the EWS region). To mathematically model this pressure behavior, a very general solution is proposed using a form
of the Arps equation. That is we propose representing the EWS pressure decline as
p(t ) ISIP ( ISIP EWS ) f (t, , )

(B.26)

where

f (t , , ) 1/ (1 (t t p ) / )1

(B.27)

Here is a power constant where 0 represents an exponential decline and 1 represents a harmonic decline. The
1 dp
constant is a pressure decline time constant (i.e.,

p dt

).
t t p

Discrete Fracture Network (far-field) Region


Shut-in Mass Conservation
The mass conservation equations during shut-in for the DFN are now addressed. After the EWS region closes, the DFN
system will begin to close. Mass conservation during shut-in requires that the DFN fracture volume during shut-in be equal to
the fracture volume at the time of the EWS closure (essentially equal to the end of pumping since the fracture efficiency is
very large) minus the fluid loss after pumping due to leakoff:
VDFN (t ) VDFN (t p ) V

DFN

(t ) VEWS (t p ) V

EWS

(tc

EWS

(B.28)

where the fluid leakoff volume from the DFN is given by


V

DFN

C (t p ) ADFN (t p ) t p G( ) 2

(B.29)

The fracture volume ratio during shut-in in the EWS region as a function of time is found by substituting Eq. B.28 into Eq.
B.29 for a constant leakoff coefficient ( C (t t p ) C (t p ) ):
VDFN (t ) VDFN (t p ) C (t p ) ADFN (t p ) t p G( ) 2 VEWS (t p ) C (t p ) AEWS (t p ) t p G(c

EWS

) 2

(B.30)

36

SPE 140514

Further assuming that the DFN volume at the time of closure of the EWS is equal to that at the end of pumping we have
VDFN (t ) VDFN (t p ) C (t p ) / 2 ADFN (t p ) t p G( ) G(c

EWS

(B.31)

or
VDFN (t )
(1 DFN )
1
G( ) G(c
VDFN (t p )
2 cDFN

EWS

(B.32)

where DFN is the fracture efficiency of the DFN which is essentially equal to the total system efficiency (since the EWS
volume is much less than the DFN volume ( VDFN VEWS )).
The resulting closure condition from Eq. B.32 is
2
G(c DFN ) c DFN G(c EWS )
(B.33)
1 DFN
Pressure Decline
Assuming the fracture compliance remains constant during closure (i.e., self-similar closure), we have
pDFN (t )
V (t )
DFN
pDFN (t p ) VDFN (t p )

(B.34)

where pDFN (t ) = p f (t ) is the fracture net pressure.


The governing fracture pressure decline equation during closure is obtained by substituting Eq. B.34 into B.33, resulting
in
pDFN (t )
(1 DFN )
1
G( ) G(c
pDFN (t p )
2 cDFN

EWS

(B.35)

Placing Eq. B.35 in terms of the ISIP (i.e., pDFN (t p ) ISIP EWS ) we find
p(t p ) p(t ) G( ) G(c

EWS

dP
dG

(B.36)

where
1 DFN
dP
pDFN (t p )

dG
2 c DFN

(B.37)

and P p(t p ) p(t ) with p(t p ) ISIP EWS .


Pressure Derivative Ratio
The pressure derivative ratio during closure of the complex fractures in the EWS region to the far-field DFN region from Eqs.
B.21 and B.35 is
dP
dG
dP
dG

EWS

DFN

VDFN (t p ) VEWS (t p )

VEWS (t p )

1 DFN
pDFN (t p )

2 c DFN

1 s
pEWS (t p )
2 c s

pEWS (t p ) VDFN (t p ) VEWS (t p )

pDFN (t p )
VEWS (t p )

(B.38)

Thus, since VDFN (t p ) VEWS (t p ) and normally pEWS (t p ) pDFN (t p ) the initial pressure decline slope when EWS is present
will be much greater than the later pressure decline slope during closure of the planar or far-field fracture network. This
analysis also predicts multiple closures and inflection points as a result of fracture complexity in the mid-field or the EWS
region.

Вам также может понравиться