Petroleum 6 (2020) 198–205

Contents lists available at ScienceDirect

Petroleum
journal homepage: http://www.keaipublishing.com/en/journals/petroleum

A tight sandstone multi-physical hydraulic fractures simulator study and its T

field application
Yonghong Wanga,b, Binshan Jua, Shihao Wangc,∗, Zhenzhou Yangb, Qing Liub
a
School of Energy Resource, China University of Geoscience, Beijing, China
b
CNPC Engineering Technology R&D Company Limited (CPET), China
c
Petroleum Engineering Department, Colorado School of Mines, Golden, CO, 80401, USA

ARTICLE INFO ABSTRACT

Keywords: During the past years, the recovery of unconventional gas formation has attracted lots of attention and achieved
Tight sandstone； huge success. To produce gas from the low-permeability unconventional formations, hydraulic fracturing
Hydraulic fracture simulator； technology is essential and critical. In this paper, we present the development of a three-dimensional thermal-
Integrated finite difference discretization； hydraulic-mechanical numerical simulator for the simulation of hydraulic fracturing operations in tight sand-
Stress contrast
stone reservoirs. Our simulator is based on integrated finite difference (IFD) method. In this method, the si-
mulation domain is subdivided into sub domains and the governing equations are integrated over a sub domain
with flux terms expressed as an integral over the sub domain boundary using the divergence theorem. Our
simulator conducts coupled thermal-hydraulic-mechanical simulation of the initiation and extension of hy-
draulic fractures. It also calculates the mass/heat transport of injected hydraulic fluids as well as proppants. Our
simulator is able to handle anisotropic formations with multiple layers. Our simulator has been validated by
comparing with an analytical solution as well as Ribeiro and Sharma model. Our model can simulate fracture
spacing effect on fracture profile when combining IFD with Discontinuous Displacement Method (DDM).

1. Introduction enable crack inside an element. In this way, XFEM resolves the diffi-
culty of remeshing. Recently, XFEM has been successfully used to model
The numerical approaches to quantify the performance of hydraulic hydraulic fracturing [11–13]. Although XFEM is very flexible in
fracturing operations have been investigated since the 1960s. Perkins handing fracture extensions, it has to discretize and solve the entire
and Kern [1], who applied the plane strain crack solution by Sneddon computational domain (formation in our case), which prevents it from
[2], developed the earliest hydraulic fracturing model, called the PK being applied in field-scale simulation, especially real-time simulation.
model. The PK model was later extended by Nordgren to PKN model Recently, Wu and Olson developed a multistage fracturing simulator
[3]. In the PKN model, the fractures are assumed to be long and with based on Discontinuous Discontinuity Method (DDM) [14–16], and
constant height and elliptical cross-section. Geertsma and De Klerk conduct Numerical Investigation of complex fracture networks in
developed the KGD model in 1969, which assumes the fractures to be naturally fractured reservoirs. DDM is essentially an extension of
short and to be under plain strain. Clearly, PKN and KGD models are Boundary Element Method (DEM), which is based on analytical solution
rough estimations of the fracture geometry. For the field applications in of point source and Green's theorem. DDM is the solution of plane
complex multi-layer reservoirs, three-dimensional models of hydraulic elasticity problems by the displacement discontinuity method [17],
fractures with fluid/proppant transport are demanding. To simulate which solves the computational domain on the boundary only, reducing
several numerical approaches have been developed. a 3D problem to a 2D one. In this sense, DDM is naturally suitable of the
Finite Element Method (FEM) has been widely used in fracture growth of hydraulic fractures with relatively less computational re-
mechanics [5–7]. Traditional FEM needs frequent remeshing to accu- sources. Li developed a fully coupled thermal-hydraulic-mechanical
rately capture the dynamic growth of the fracture. Extended Finite hydraulic-fracturing simulator based on mixed finite-volume/finite
Element Method (XFEM) resolves this issue [8–10]. XFEM adopts an element method [18]. The simulator has been used to simulate multi-
‘enriched’ base function and the Partition of Unity Method (PUM) to stage fracturing and has achieved good performance. However, the

Peer review under responsibility of Southwest Petroleum University.

∗
Corresponding author.
E-mail address: [email protected] (S. Wang).

https://doi.org/10.1016/j.petlm.2019.05.001
Received 4 December 2018; Received in revised form 29 April 2019; Accepted 23 May 2019
2405-6561/ Copyright © 2020 Southwest Petroleum University. Production and hosting by Elsevier B. V. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
Y. Wang, et al. Petroleum 6 (2020) 198–205

model does not account for the proppant transport in the fracture. Wu KI
KIC
hx
and Olson have extended DDM to simulate the interaction between d x = vfront t , .......KI > KIC
min H (x )
hydraulic fractures and natural fractures [19]. Later, Tang combined KIC +
hz (5)
DDM with Finite Difference Method to account for the transport of
fluid/proppant in the fracture systems in 2016 [20]. Prashanth opti- KI
KIC
hz
mize multiple fractures propagating simultaneously in hydraulic frac- d z = vfront t
min H (x )
, .......KI > KIC
turing to achieve uniform growth using data-based model reduction in KIC +
hz (6)
2018 [21]. Guo develop a Reservoir-Geomechanics-Fracturing Model to
analyze the Effects of Subsequent Parent Well Injection on Interwell In the above equation, vfront is the maximum velocity at which the
fracture front extends. t is the time step length. h x and hz is the
Fracturing Interference in 2018 [22]. Yu develop a numerical model to
dimension of the grid block along x and z-direction respectively, H (x ) is
simulate pressure response of well interference and well performance in
the fracture height, varying along the fracture extension direction (x-
tight oil reservoir by fast embedded-discrete-fracture-model (EDFM)
direction). hz is the position of the grid block along the vertical direc-
method in 2018, the fracture model is complex fracture geometries.
tion (z-direction) in the fracture plain.
And simulate shale gas transport and production with complex fractures
model [23,24].
In this paper, we have developed a multi-physical hydraulic frac- 2.2. Mass/energy transport inside the fracture
tures simulator for tight sandstone, and solve fractures model by
Integrated Finite Difference Discretization (IFD), which is easy to get The injected slurry is assumed to be a mixture of multiple species,
convergent solution and save calculation time based on our analyses. including solid species (proppants) and fluid species (water/gel/ad-
Our simulator can calculate the mass/heat transport of injected hy- ditives). The fracture is simulated on a pre-set grid. The dimension
draulic fluids as well as proppants, and handle anisotropic formations along which the fracture extends is set as the x-direction. The fracture
with multiple layers, analyze different mudstone/shale layer, stress grid blocks are allowed to deviate along the direction that is perpen-
contrast and fractures spacing effect for fractures extension. dicular to the x-direction. The deviation direction is set as the y-di-
rection. The vertical direction is set as the z-direction.
The general mass balance equation inside the fracture is
2. Mathematical model
+ ( v)+q=0
2.1. Fracture mechanics t (7)
where, v is velocity, is the density and q is the sink/source term.
As we know, the injection of fracturing fluids builds up the down- We consider the fracture to be a thin slit in the Cartesian y-direction
hole pressure and consequently causes the formation to fail. There are and integrate Equation (5) over that width to yield:
several models to describe the failure of rock, namely Mode-1 failure
(tearing), Mode-2 failure (shearing) and Mode-3 failure (hybrid model). ( w)
( v w) + ( v w) + +q=0
At the time being, we only consider Mode-1 failure, the stress intensity x z t (8)
factor at the fracture tip is as shown in Equation (1), [25]. where w is the fracture width.
1/2 Slurry fluid and proppant components are conserved. A version of
E 2
KI = wfr (r ) Equation (6) for a fluid component is obtained by defining the fluid
8(1 2) r (1)
mass per unit volume as the product of fluid density and fluid compo-
where KI is the stress intensity factor at the fracture tip and r is the nent volume fraction in the overall fluid volume, and defining the fluid
inward normal distance from the fracture front. is the Poisson's ratio. velocity as the fluid superficial velocity averaged over the fracture
wfr is the width of the fracture front. width. The source/sink terms consist of fluid leakoff, the flow of fluid
The velocity at which the fracture front extension can be calculated through the porous fracture wall, and flow of fluid between the com-
as [26]: pleted zones along the wellbore (the surroundings) and the fracture.
Then:
n
KI KIC
vfront = Cext
(2)
Np Np
KIC
f xf 1 cpi vfl w + f xf 1 cpi vfl w
x z
where Cext and n are parameters that rock properties and KIC is the rock i=1 i=1

toughness.
Np
The fracture width is dependent on the net pressurePnet , which is the
+ xf 1 cpi w + x f qleak = 0
difference between the fracture pressure Pf induced by the injected t f
i=1
f

slurry and the minimum principal stress min , as (9)

Pnet = Pf min (3) In the above equation, qleak is the leakoff volume. Np is the number of
proppant types. cpi is the proppant concentration of the ith type of
The fracture equation is derived from the integral of the deflection proppant.
induced by a point-source load over the boundary of a semi-infinite A version of Equation (6) for a proppant component is obtained by
medium [27]. defining the proppant mass per unit volume as the product of proppant
2) grain density and proppant volume fraction, and defining the proppant
(1
w(r ) = Pnet dA velocity as the proppant superficial velocity averaged over the fracture
Er (4)
A width. Proppant velocity does vary for each proppant component and
does not leak off through the fracture face, so the leakoff term in
where w is the fracture width, and A is the area the load acts on. r
Equation (7) is absent. Then:
denotes the coordinates of the point of interest. The fracture front ex-
tends when the stress intensify factor at the front is larger than rock i i
i i i i i i
failure toughness. The extension distance along the horizontal direction p cp v p w + p cp v p w + p cp w =0
x z t (10)
(x-direction) and the vertical direction (z-direction) is calculated as
shown in Equation (5) and Equation (6) respectively. Proppant is assumed to be incompressible. Because a fracture

199
Y. Wang, et al. Petroleum 6 (2020) 198–205

occupies a relatively small range of depth, we consider fluid component 0.3736µfl0.57

i
density in Equation (8) as constant. Then, Equations 7 and 8 become: f (NRe )= 0.29
0.29 i 0.86
fl p fl (dpi)
Np Np (20)
xf 1 cpi vfl w + xf 1 cp _ i vfl w
x i=1 z i=1 Np
3
Np
2
Np
gp = 5.9 cpi + 8.8 cpi 4.8 cpi + 1
Np
i=1 i=1 i=1 (21)
+ xf 1 cpi w + x f qleak = 0
t i=1
(11)
2
dpi dpi
h (w ) = 0.563 1.563 +1
And w w (22)
i i For the direction perpendicular to the gravity vector, the proppant
(cpi v p w ) + (cpi v p w ) + (cpi w ) = 0
x y t (12) velocity is effected by the fracture wall and other proppant particles.
Friehauf modified this velocity by introducing a multiplicative re-
If we sum Equation (9) over fluid components, Equation (10) over
tardation factor for the slurry velocity that accounts for both of these
proppant components, and add the results together, we obtain an
[28]:
overall slurry conservation equation:
vpi , x = Cret (cpi, w ) vsli , x (23)
Np Np Np Np

1 cpi vf + cpi v
i
p w+ 1 cpi vf +
i
cpi v p w This multiplicative factor is given by:
x i=1 i=1 z i=1 i=1 2
dpi dpi
w Cret (cpi, w ) = 1 + 2.02
+ + qleak = 0 wc wc (24)
t (13)

We equate the coefficient of w in the length derivatives to the slurry And

velocity, a weighted sum of fluid and proppant velocities: Np
0.8
1 1 1
= 1.411 cpi
Np Np wc2 (dpi)
2
w2
vsl = 1 cpi vfl + cpi v p
i i=1 (25)
i=1 i=1 (14) Fluid velocity is then obtained from slurry velocity and proppant
velocity by rearranging Equation (12):
Equation (11) then becomes:
Np i i
w vsl c v
i=1 p p
( vsl w ) + ( vsl w ) + + qleak = 0 vfl =
x z t (15) (1
Np i
c) (26)
i=1 p

The slurry velocity through the fracture is The slurry density sl is calculated by weighting the density of the
w2 fluid components and the solid components, as
vsl = (Pf + sl z )
12µsl (16) Np Np Nf
= cpi + 1 cpi x fi
where γsl is the gravity term. The approach to calculate the slurry sl pi pi fi

viscosity will be described in the section below. Proppant component

i=1 i=1 i=1 (27)
and fluid velocities generally differ because proppant is denser than In the above equation, i refers to the ith component. p and f refers to
fluid and consists of discrete granules that can interact with each other proppant (solid) and fluid components, respectively. Np and Nf is the
and with the fracture wall. Therefore, the z-direction velocity of prop- number of proppant species and number of fluid species respectively. c
pant is the sum of the slurry velocity plus a proppant settling velocity and x is the volume fraction of proppant component and fluid respec-
that accounts for the difference in fluid and proppant component den- tively.
sities, as
3. Integrated finite difference discretization numerical approach
vpi , z = vsl + vpi , stl (17)
where vpi , z is proppant component velocity and vpi , stl is proppant com- Our simulator is based on uniform grid. The mass/heat transport
ponent settling velocity. Proppant settling velocity is calculated based governing equation is solved in a 2D plane, while the mechanical im-
on Stoke's law as pact is enforced on the third direction that is normal to the grid plane.
Therefore, our simulator conducts 3D simulation of the fracturing
i
g (dpi)
2 process. The grid block (Fig. 1) can be assigned with different proper-
p fl
ties to account for the heterogeneity of the formation. Our simulator
vpi , stokes =
18µ fl (18) solves the governing equations by integrated finite difference (IFD)
method, as proposed by Narasimhan and Witherspoon in 1976 [29].
where is the Stoke's velocity and
vpi , stokes is the proppant grain dia- dpi In this method, the simulation domain is subdivided into sub do-
meter of the ith proppant respectively. Our simulator adopts Friehauf's
mains and the governing equations are integrated over a sub domain
approach to calculate the settling velocity [28], the proppant settling
with flux terms expressed as an integral over the sub domain boundary
velocity is the following form:
using the divergence theorem. The definitions of the geometric para-
vpi , stl = vpi , stokes f (NRe
i
) gp h (w ) meters used in the following derivation of this method. We consider a
(19)
generalized conservation equation of the form:
where,NRe
i
is Reynold's number, f captures inertial effects, gp i
(NRe )
Mk
captures the effect of interfering proppant particles, and h (w ) captures = F k + qk
t (28)
the effect of the fracture wall. These expressions, modified for the
presence of multiple proppant components, are: where k refers to conserved quantity, M is quantity per unit volume, F is

200
Y. Wang, et al. Petroleum 6 (2020) 198–205

Fig. 1. Concpetual model of fracture on the grid. Green grid blocks refer to
newly activated (cracked) grid blocks. Blue grid blocks refer to grid blocks that Fig. 4. The profile of the multi-layer formation.
are cracked in previous time steps.

Table 1
Input parameters for the radial fracture case.
Properties Values Units

Young's modulus 17.24 GPa

Poisson's ratio 0.25 dimensionless
Biot's coefficient 1.0 dimensionless
Grid block length (x-direction) 10 m
Grid block height (z-direction) 10 m
Injection rate 3.18 m3/min
Injection time 30 min
Frac fluid viscosity 40 cp

Fig. 2. Conceptual model of integral finite difference method [29,30].

where l is time level.
flux, and q is source/sink term. Integrating Equation (28) over a sub These governing equations (Equation (32)) expressed in residual
domain Vn yields: vector form are:

R (xl + 1) = 0 (33)
M kdV = F k nd
ˆ + qkdV
t Vn n Vn (29)
where, xl + 1 is the primary variable vector at time level l+1.
Integrals over Vn are replaced with averages: The time step length t is calculated based on the
M kdV = Mnk Vn Courant–Friedrichs–Lewy (CFL) number as
Vn (30)
vinj t
Integrals over the sub domain boundary are replaced with discrete CFL =
sums over boundary-averaged segments:
min( x , z ) (34)

k k where vinj is the velocity of the fluid in the wellbore, calculated from the
ˆ =
F nd Anm F nm
n (31) wellbore model. x and z are the grid block length along the x and z
m
direction, respectively. To ensure the convergence, the CFL number
where subscript n denotes an averaged quantity over Vn, Anm is the should not exceed 1. In this work, it is set to be 0.8 by default.
boundary segment common to Vn and Vm, double subscript nm denotes Fractures extension depend on formation stress field, this stress field
an averaged quantity over boundary segment Anm. The time derivative will change when a fracture appears by hydraulic fracturing, so the
is approximated by the standard first order finite difference approx- follow-up fractures profile will also change, this phenomena is called
imation. Applying these approximations to Equation (29) yields: stress shadow. We can combine the IDF with Discontinuous
Displacement Method (DDM) to describe the stress shadow effect and
l+1 l t
[Mnk] [Mnk] k
Anm Fnm + Vn qnk = 0 simulate different stages fractures. DDM is essentially a boundary ele-
Vn (32)
m ment type method, it can calculate the induced stress field, this

Fig. 3. Simulated fracture width after 30 min injection.

201
Y. Wang, et al. Petroleum 6 (2020) 198–205

Table 2
Three cases for different mudstone layer thickness.
case upper mudstone sandstone lower mudstone

Thickness,m Stress,MPa Thickness,m Stress,MPa Thickness,m Stress,MPa

1 15 58 12 52 15 58
2 20 58 12 52 20 58
3 25 58 12 52 25 58

Fig. 5. The fracture profile for case 1 (mudstone thickness is 15 m).

Fig. 6. The fracture profile for case 2 (mudstone thickness is 20 m).

Fig. 7. The fracture profile for case 3 (mudstone thickness is 25 m).

calculated stress field is added to grid blocks in IFD to determine dis-

placement, and the vertical displacement can be defined as fracture
width. Tang described the method of how to combine the DDM and
another fracture model [20].

4. Validation

We first compare the radial fracture extension problem is a simu-

lation of fracture extension in the horizontal plane with no leakoff,
gravity, or temperature effects [31] (see Fig. 2). Fig. 3 shows simulated
Fig. 8. Proppant distribution strength (proppant mass/area) at near wellbore fracture width after 30 min of injection. The fracture geometry for both
fracture profile for case 3. models are similar, the fracture width profiles from both simulators are
very similar. The fracture half-length from our simulator is about

202
Y. Wang, et al. Petroleum 6 (2020) 198–205

Table 3
Three cases for different mudstone stress.
case Upper mudstone sandstone Lower mudstone

Thickness,m Stress,MPa Thickness,m Stress,MPa Thickness,m Stress,MPa

1 15 58 12 52 15 58
2 15 62 12 52 15 62
3 15 72 12 52 15 62

Fig. 9. The fracture profile for case 2 (mudstone stress is 62 MPa).

Fig. 10. The fracture profile for case 3 (mudstone stress is 72 MPa).

with sandstone layers and mudstone layers appearing in sequence. The

mudstone layers play like interlayers in this case study. The perforation
(injection point) locates at the sandstone layer in the middle. The
conceptual model of the multi-layer formation and the position of
perforation is shown in Fig. 4.
Some input parameters as follows, formation pressure is 24 MPa,
and formation temperature is 90 °C, perforation layer is between 3269
and 3273 m, frac fluid is injected through tubing, tubing ID is
61.98 mm, injection rate is 2.4 m3/min, total injection volume is
180 m3, average proppant concentration is 400 kg/m3, total proppant is
49.2t. The Young's module for sandstone is 41 GPa, the Young's module
Fig. 11. Proppant distribution strength (proppant mass/area) at near wellbore
for mudstone is 48 GPa, and Poisson's ratio for sandstone and mudstone
fracture profile for case 3.
are 0.22 (see Table 1).

94.5 m (310 ft), the fracture half-length from Ribeiro and Sharma is
100.5 m (330 ft), they are also similar. Ribeiro and Sharma also com- 5.1.1. Mudstone layer thickness effect for fractures extension
pared fracture radius versus time to an analytical solution for radial We assume three cases (in Table 2), mudstone thickness is 15 m,
fractures with no leakoff from Geertsma and obtained good agreement 20 m and 25 m. Sandstone stress is 52 MPa, thickness is 12 m, and
[4,31]. mudstone stress is 58 MPa, so the stress contrast between sandstone and
mudstone is 6 MPa. We simulate fractures profile for three cases by our
model.
5. Field case study From Figs. 5–7, comparing fracture profiles of three cases, it shows
that as the thickness of mudstone layer increase, the resistance of
5.1. The mudstone interlayers effect on fractures extension fracture upward extension increase, and it is easier to form a long and
wide fracture. fracture length is from 230 m to 310 m, the maximum
In this section, we present a series of case study to investigate the fracture width is from 1.3 cm to 1.5 cm, and fracture extension is
impact of thin layers with stress contrast on fracture growth as well as mainly restricted in sandstone, fracture cannot extend to upper and
proppant distribution. We run our simulator on a multi-layer reservoir lower mudstone. Propant transport only in sandstone if fracture cannot

203
Y. Wang, et al. Petroleum 6 (2020) 198–205

Fig. 12. Three dimension fractures profile at spacing 20 m and 40 m

Fig. 13. Map view of fractures profile at spacing 20 m and 40 m

extend to mudstone, so fractures cannot be the channel between ratio is 0.22, injection rate is 3 m3/min, injection time is 60mins,
sandstone and upper/lower sandstones. On the other hand, from Fig. 8, proppant density is 2650 kg/m3, frac fluid viscosity is 170cp, frac fluid
although the fracture penetrates the layer above and below, not many density is 1100 kg/m3, and simulate spacing is 10 m and 40 m. Fig. 12
proppants enter the sandstone layers above and below. In this sense, and Fig. 13 show the 3D fractures profile and map view of fractures
when the pumping stops and the fracture close, there will be no ef- profile, the effect of fracture spacing on fracture profile is very obvious,
fective (propped) fracture in the sandstone layers above and below. the smaller the fracture spacing, the more obvious the fracture bending,
This phenomenon is partly because the fracture toughness of mudstone the larger the fracture spacing, the more uniform the fracture extension.
and tight sandstone in this reservoir is close to each other. The fractures profile is bending when fracture spacing is 20 m, and the
fractures profile is uniform when fractures spacing increase to 40 m.
5.1.2. Stress contrast between sandstone and mudstone effect for fracture
6. Summary
extension
We also assume three cases (in Table 3), mudstone stress is
In this paper, we have developed a practical multi-physical hy-
58 MPa、62 MPa、72 MPa. Sandstone thickness is 12 m, mudstone
draulic fracturing simulator for tight sandstone reservoirs.
thickness is 15 m, sandstone stress is 52 MPa, so the stress contrast
between sandstone and mudstone is 6 MPa、10 MPa、20 MPa. We si-
(1) The simulator is able to accurately capture the initiation and ex-
mulate fractures profile for three cases by our model.
tension of fractures in complex multi-layer reservoirs. The accuracy
The fracture profile for case 1 is Fig. 5. From Figs. 5 and 9 and
of our simulator has been validated by comparing with an analy-
Fig. 10, we can find that as mudstone stress increase, the resistance for
tical solution as well as Ribeiro and Sharma model.
fracture extension increase obviously, and fracture length increase ob-
(2) In the reservoir of the Sulige gas field, the minimum principal stress
viously from 230 m to 320 m. In case 3 where the minimum principal
difference between shale and sandstone is 5～9 MPa. According to
stress of the upper mudstone layer further increases to 72 MPa, the
our simulation, the fracturing operation will open the shale layer
upper layer cannot be penetrated. Due to the increase of the formation
sand and make channels between the upper and the lower sand-
stress, the fracture width decreases while the fracture length increases.
stone layer. However, proppants cannot always enter the fracked
Meanwhile, a significant amount of proppants settles downward to the
layers. In this way, gel with stronger proppant-carrying capability
lower sandstone layer. In this way, the lower sandstone layer is also
should be adopted in such formations. On the other hand, wells
propped (see Fig. 11).
with fractures that grow into multiple layers may have potential
water conning issues, the water from the lower layers may enter the
5.2. Fractures spacing effect on fractures profile wellbore through the hydraulic fractures at the later stage of pro-
duction. Regarding this issue, cautions must be taken to frack
In this section, we assume four fractures extend simultaneously, water-bearing formations, especially when a water-bearing layer is
formation is homogeneous, rock Young's module is 41 GPa, Poisson

204
Y. Wang, et al. Petroleum 6 (2020) 198–205

below the pay zone and the formation stress contrast is not high org/10.2118/167626-PA.
enough to prevent the penetration of the fracture. [15] K. Wu, J.E. Olson, Mechanics analysis of interaction between hydraulic and natural
fractures in shale reservoirs, Unconventional Resources Technology Conference
(3) Fracture spacing affect fracture profile obviously, and the smaller (URTEC). Denver, CO, 2014 https://doi.org/10.15530/URTEC-2014-1922946.
the fracture spacing, the more obvious the fracture bending, the [16] K. Wu, J.E. Olson, Investigation of the impact of fracture spacing and fluid prop-
larger the fracture spacing, the more uniform the fracture exten- erties for interfering simultaneously or sequentially generated hydraulic fractures,
SPE Prod. Oper. 28 (2013) 427–436 https://doi.org/10.2118/163821-PA.
sion. [17] S.L. Crouch, Solution of plane elasticity problems by the displacement discontinuity
method. I. Infinite body solution, Int. J. Numer. Methods Eng. 10 (1976) 301–343.
References [18] S. Li, X. Li, D. Zhang, A fully coupled thermo-hydro-mechanical, three-dimensional
model for hydraulic stimulation treatments, J. Nat. Gas Sci. Eng. 34 (2016) 64–84
https://doi.org/10.1016/J.JNGSE.2016.06.046.
[1] T.K. Perkins, L.R. Kern, Widths of hydraulic fractures, J. Pet. Technol. 13 (1961) [19] K. Wu, J.E. Olson, M.T. Balhoff, W. Yu, Numerical analysis of promoting more-
937–949 https://doi.org/10.2118/89-PA. uniform development of simultaneous multiple fracture propagation in horizontal
[2] I.N. Sneddon, The distribution of stress in the neighbourhood of a crack in an elastic wells, SPE Prod. Oper. 32 (01) (2017).
solid, Proc. R. Soc. A Math. Phys. Eng. Sci. 187 (1946) 229–260. [20] H. Tang, P.H. Winterfeld, Y.-S. Wu, Z. Huang, Y. Di, Z. Pan, J. Zhang, Integrated
[3] R.P. Nordgren, Propagation of a vertical hydraulic fracture, Soc. Petrol. Eng. J. 12 simulation of multi-stage hydraulic fracturing in unconventional reservoirs, J. Nat.
(1972) 306–314 https://doi.org/10.2118/3009-PA. Gas Sci. Eng. 36 (2016) 875–892 https://doi.org/10.1016/j.jngse.2016.11.018.
[4] J. Geertsma, F. De Klerk, A rapid method of predicting width and extent of hy- [21] S. Prashanth, K. Wu, K. Joseph, Optimizaiton of simultaneously propagating mul-
draulically induced fractures, J. Pet. Technol. 21 (1969) 1571–1581 https://doi. tiple fractures in hydraulic fracturing to achieve uniform growth using data-based
org/10.2118/2458-PA. model reduction, Chem. Eng. Res. Des. 136 (2018) 675–686.
[5] S.H. Advani, T.S. Lee, J.K. Lee, Three-dimensional modeling of hydraulic fractures [22] X. Guo, K. Wu, C. An, J. Tang, J. Killough, Numerical investigation of effects of
in layered media: Part I—finite element formulations, J. Energy Resour. Technol. subsequent parent well injection on Interwell fracturing interference using re-
112 (1990) 1 https://doi.org/10.1115/1.2905706. servoir-geomechanics-fracturing modeling, SPE J. (2018) accepted https://doi.org/
[6] R.G. Baca, R.C. Arnett, D.W. Langford, Modelling fluid flow in fractured-porous 10.2118/195580-PA.
rock masses by finite-element techniques, Int. J. Numer. Methods Fluids 4 (1984) [23] W. Yu, Y. Xu, R. Weijermars, K. Wu, K. Sepehrnoori, A numerical model for si-
337–348 https://doi.org/10.1002/fld.1650040404. mulating pressure response of well interference and well performance in tight oil
[7] Z. Chen, Finite element modelling of viscosity-dominated hydraulic fractures, J. Pet. reservoirs with complex-fracture geometries using the fast embedded-discrete-
Sci. Eng. 88–89 (2012) 136–144 https://doi.org/10.1016/J.PETROL.2011.12.021. fracture-model method, SPE Reservoir Eval. Eng. 21 (02) (2018) 489–502 https://
[8] N. Moes, T. Belytschko, Extended finite element method for cohesive crack growth, doi.org/10.2118/184825-PA.
Eng. Fract. Mech. 69 (2002) 813–833 https://doi.org/10.1016/S0013-7944(01) [24] W. Yu, Y. Xu, M. Liu, K. Wu, K. Sepehrnoori, Simulation of shale gas transport and
00128-X. production with complex fractures using embedded discrete fracture model, AIChE
[9] N. Moes, J. Dolbow, T. Belytschko, A finite element method for crack growth J. 64 (06) (2018) 2251–2264.
without remeshing, Int. J. Numer. Methods Eng. 46 (1999) 131–150 https://doi. [25] C.H. Yew, X. Weng, Mechanics of Hydraulic Fracturing, second ed., Gulf
org/10.1002/(SICI)1097-0207(19990910)46:1<131::AID-NME726>3.0.CO;2-J. Professional Publishing, Waltham, MA, 2015.
[10] N. Sukumar, N. Moes, B. Moran, T. Belytschko, Extended finite element method for [26] E.N. Mastrojannis, L.M. Keer, T. Mura, Growth of planar cracks induced by hy-
three-dimensional crack modelling, Int. J. Numer. Methods Eng. 48 (2000) draulic fracturing, Int. J. Numer. Methods Eng. 15 (1980) 41–54 https://doi.org/
1549–1570 https://doi.org/10.1002/1097-0207(20000820)48:11<1549::AID- 10.1002/nme.1620150105.
NME955>3.0.CO;2-A. [27] A.E.H. Love, E.H. Augustus, A Treatise on the Mathematical Theory of Elasticity,
[11] B. Lecampion, An extended finite element method for hydraulic fracture problems, Cambridge University Press, 2013.
Commun. Numer. Methods Eng. 25 (2009) 121–133 https://doi.org/10.1002/cnm. [28] K.E. Friehauf, Simulation and Design of Energized Hydraulic Fractures, The
1111. University of Texas at Austin, 2009.
[12] T. Mohammadnejad, A.R. Khoei, An extended finite element method for hydraulic [29] T.N. Narasimhan, P.A. Witherspoon, An integrated finite difference method for
fracture propagation in deformable porous media with the cohesive crack model, analyzing fluid flow in porous media, Water Resour. Res. 12 (1976) 57–64 https://
Finite Elem. Anal. Des. 73 (2013) 77–95 https://doi.org/10.1016/J.FINEL.2013.05. doi.org/10.1029/WR012i001p00057.
005. [30] K. Pruess, A practical method for modeling fluid and heat flow in fractured porous
[13] S. Salimzadeh, N. Khalili, A three-phase XFEM model for hydraulic fracturing with media, Soc. Petrol. Eng. J. 25 (1985) 14–26 https://doi.org/10.2118/10509-PA.
cohesive crack propagation, Comput. Geotech. 69 (2015) 82–92 https://doi.org/10. [31] L. Ribeiro, M.M. Sharma, A new three-dimensional, compositional, model for hy-
1016/J.COMPGEO.2015.05.001. draulic fracturing with energized fluids, SPE Annual Technical Conference and
[14] K. Wu, J.E. Olson, Simultaneous multifracture treatments: fully coupled fluid flow Exhibition. Soc. Petrol. Eng, 2012.
and fracture mechanics for horizontal wells, SPE J. 20 (2015) 337–346 https://doi.

205