Applied Energy 183 (2016) 1664–1681

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

A simplified model for heat extraction by circulating fluid through a

closed-loop multiple-fracture enhanced geothermal system
Bisheng Wu a,⇑, Xi Zhang a, Robert G. Jeffrey b, Andrew P. Bunger c,d, Shanpo Jia e
a
CSIRO Energy, Melbourne, VIC, Australia
b
SCT Operations Pty Ltd, Wollongong, NSW, Australia
c
Department of Civil and Environmental Engineering, University of Pittsburgh, Pittsburgh, PA, USA
d
Department of Chemical and Petroleum Engineering, University of Pittsburgh, Pittsburgh, PA, USA
e
Research Center of Geomechanics and Geotechnical Engineering, Yangtze University, Jingzhou, Hubei, China

h i g h l i g h t s

Heat extraction from a multiple fracture system with dipole wells is considered.

The controlling parameters are found via a scaling analysis.

Semi-analytical solutions are obtained via Laplace transform and conjugate function.

Optimal values for the number of fractures and fracture spacing are found.

The present model works as an efficient and effective tool for EGS design.

a r t i c l e i n f o a b s t r a c t

Article history: Multiple hydraulic fractures have been proposed for improving the performance of an enhanced geother-
Received 19 April 2016 mal system (EGS) by providing conductive flow pathways and increased contact area between flowing
Received in revised form 16 September fluid and surrounding rock formation. Use of more fractures incurs a higher drilling and hydraulic frac-
2016
turing cost, but the additional cost can be offset by improved operation performance of an EGS. In this
Accepted 28 September 2016
Available online 7 October 2016
paper, a model is presented for efficiently predicting the output temperature so as to optimize the num-
ber of fractures and fracture spacing to maximize the EGS lifetime under a constant circulation rate. This
optimal spacing is shown to arise due to the interplay among number of fractures, fracture spacing, well
Keywords:
Enhanced geothermal system
depth, and the pre-existing geothermal gradient. Specifically, under a typical geothermal gradient asso-
Multiple parallel fractures ciated with EGS for a 5 km total vertical depth of the well, the number of fractures N and the equal frac-
Semi-analytical solutions ture spacing d have optimal values: 6 6 N 6 13 and 30 m 6 d 6 90 m. In addition, the semi-analytical
Runge-Kutta method solution method presented is effective and efficient in computation and, for this reason, is useful for opti-
Scaling analysis mizing the design of a geothermal reservoir with multiple layers at equal or non-equal spacing.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction 12,729 MWe worldwide as of 2015 [5]. According to Geodynamics

Ltd. Australia, the measured temperature of the granite geothermal
Geothermal energy is one of most promising alternatives for reservoir in their Joloka 1 geothermal well ranges from 273 to
providing future renewable energy supplies because it is a nearly 283 °C at a depth of approximately 5 km and the estimated
inexhaustible underground resource that can be exploited using geothermal resource is equivalent to the electrical energy that
low emission methods [1]. In addition to its wide use for district would be provided by burning 16 trillion cubic feet of natural
heating by using ground source heat pumps (GSHP) [2–4], geother- gas [6]. Due to low flow permeability of the rock mass and slow
mal energy can also be used for power generation from conven- natural heat convection [7], it is a very challenging task to extract
tional geothermal systems and enhanced geothermal systems such a large amount of heat from the earth. In order to improve the
(EGS). Conventional geothermal has a total installed capacity of output performance of an EGS and make a sustainable utilization of
the geothermal resources, multiple fractures have been proposed
⇑ Corresponding author. to provide several conductivity paths between the wells with
E-mail address: Bisheng.wu@csiro.au (B. Wu). increased contacting area with surrounding rock; however, the

http://dx.doi.org/10.1016/j.apenergy.2016.09.113
0306-2619/Ó 2016 Elsevier Ltd. All rights reserved.
 B. Wu et al. / Applied Energy 183 (2016) 1664–1681 1665

Nomenclature1

A0 geothermal gradient v2 = (cos hU  cos W)2

A‘ (‘ = 1, N) undetermined coefficients V known vector
A, B unknown vectors W known matrix
A controlling parameter Xi,Yi, Zi coordinates (i = f is for fluid flow and i = T for heat
B0 ground temperature flow)
B‘ (‘ = 1, N) undetermined coefficients X, Y, Z dimensionless coordinates
cr, cw specific heat capacity of rock and fluid x, y, z dimensional coordinates
C, D, E, F, M known matrice x‘i ; y‘i coordinates of the ith well in the ‘th fracture
dij, d distance between ith and jth fractures z‘0 dimensional vertical coordinate of ‘th fracture
DL fluid dispersion coefficient Z ‘0 dimensionless vertical coordinate of ‘th fracture
D controlling parameter
Gi (i = 1, 11) group parameters Greek letters
g(U, W) function related to the travel time r2 laplace operator
H depth of the bottom fracture x, x‘ fracture conductivity
ki (i = 1.p. ffiffi.4) vectors related to RK method kr thermal conductivity of rock
K ¼ 2v s M, known matrix jr = kr/qrcr thermal diffusivity of rock
K controlling parameter H0 = cZ, dimensionless initial temp
L half separation of wells H‘in dimensionless injection temperature for the ‘th
Lf length scale for fluid flow fracture
LT length scale for heat flow H‘r ; H‘f dimensionless temps. for the ‘th rock layer and
M number of wells fracture
M known matrix v parameter related to convection
N number of fractures v1 parameter related to time derivative
NC number of streamlines /‘f ; /f velocity potential
Q ‘f ; Q f dimensionless discharge vector U dimensionless velocity potential
Q fx, Q fy dimensionless fluid discharge W dimensionless streamline function
q‘f ; qf dimensional fluid discharge vector Wi (i = 1. . . NC + 1), Value of ith streamline
Q ‘i flow rate at the ith well a = d/H pffiffi
Q in, Q out injection and production rates b‘ (‘ = 1, N + 1) = expð2 sZ ‘0 Þ
q⁄ discharge scale r‘ = b‘+1/b‘
rw wellbore radius g = A0H/(A0H + B0)
S complex number c = ga
T ‘f ; T ‘r temperature in the ‘th fracture and layer q variable replacing U for computing, q = exp(U)
T ‘in injection temperature for the ‘th fracture qw, qr mass densities of fluid and rock
T ‘out output temperature for the ‘th fracture Dq spatial step along a specific streamline
T r ; T f temperature scales e small parameter
T0 initial temperature U⁄ velocity potential scale
T time sh travel time
U vector defined in Eq. (34) s dimensionless time
v‘f ; vf fluid velocity vector

related cost is also an important factor to be considered in such and heat transfer through a single horizontal channel with perme-
projects [8]. Therefore, an efficient numerical model is necessary able walls. In addition, Yang and Yeh [13] studied the case where
to predict the EGS performance, as it can be used to study the there is one production well and the fluid flow is radial. As the fluid
trade-off between stimulation parameters, heat extraction effi- flow is unidirectional or symmetrical within the fractures in these
ciency, and cost. models, analytical or exact solutions are obtained by using Laplace
The present work considers the thermal evolution and interac- transformation or perturbation analysis. When it comes to the
tions associated with circulation of fluid through a system of mul- dipole-well configuration (where the fluid pumped into the injection
tiple fractures, accounting for the interplay among number of well flows in the fracture(s) absorbing heat from surrounding rock
fractures, fracture spacing, well depth, and the pre-existing formation, and then is pumped out at the production well), a special
geothermal gradient. These factors are varied to find the optimal treatment i.e. velocity potential and streamline functions, is utilized.
values to extend the period of heat extraction. The context of this For example, Gringarten [14], Schulz [15] and Wu et al. [16] applied
work is provided by a number of past contributions. For the single this method to solve analytically the EGS with dipole or multiple
fracture case, exact or analytical solutions can be found when the wells. As the above models provide exact solutions in time space
fluid flow is in steady state or the geometry is simple. For example, or analytical solutions in Laplace space, they are accurate as well
Lauwerier [9] and Bödvarsson [10] studied the transport of heat as very quick in terms of computation. In addition, these solutions
during fluid flow in a straight fracture with a constant injection rate don’t have time- or space-scale restrictions.
and constant or variable injection temperature; Cheng et al. [11] For an EGS with multiple fractures or complicated geometries,
investigated in detail the effects of heat storage and dispersion on the hydrothermal coupling (sometimes mechanical deformation
the temperature evolution; Mohais et al. [12] analyzed the fluid flow is included) is complicated, leading to few analytical solutions,
and generally requires use of numerical methods. For example,
1
Gringarten et al. [17] studied a multiple parallel vertical fracture
Note: variables with a cap ‘^’ are not listed, and denote the LTs of the
EGS based on an assumption of an infinite fracture array; Fox
corresponding variables.
1666 B. Wu et al. / Applied Energy 183 (2016) 1664–1681

et al. [18] and Wu et al. [19] obtained analytical solutions for a fracture EGS with dipole wells, as well as to find the optimal values
multiple parallel fracture case by using Green’s functions and per- for the geometrical parameters. The present model is found to run
turbation methods, respectively. In the above three cases, the fluid in terms of seconds on a personal laptop computer, saving a great
flows in one dimension with uniform velocity. When there exist a deal of time compared to BEM and FEM, thus providing a good
dipole or more wells in the reservoir at the same time, as men- design tool for engineers, especially for optimization studies where
tioned above, the fluid flow is relatively complicated and thus many simulations must be carried out. On the other hand, the pre-
the problem become more difficult to solve. A wide range of open sent semi-analytical model can also be used as a benchmark for
source or commercial software are available for modelling geother- other numerical methods.
mal stimulation. To name a few here, TOUGH2 by Lawrence Berke- In addition, the present solution accounts for the effect of the
ley National Laboratory [20] uses the finite difference method geothermal gradient on the output performance and this effect is
(FDM), FEFLOW by Diersch [21] and FEHM by Los Alamos National shown below in this paper to be significant because the geother-
Laboratory [22] are based on the finite element method (FEM), and mal reservoirs normally extend over a relatively large depth
ECLIPSE by Schlumberger [23] and Fluent by ANSYS [24] use the interval.
finite volume method (FVM). They have been used successfully This paper is organized as follows. In Section 2, we will describe
to model the heat extraction from EGS reservoirs with complicated the problem studied, which is followed in Section 3 by a scaling
geometries [22,24–32]. In addition, Ghassemi et al. [33] and Kumar analysis for finding the controlling evolution parameters. The
and Gutierrez [34] used the boundary element method (BEM) to dimensionless formulation of governing equations based on fixed
predict three dimensional effects; Safari and Ghassemi [35] used length scales, more suitable to finding problem solutions, is given
one variation of BEM, i.e. displacement discontinuity element in Section 4. Section 5 provides semi-analytical solutions by com-
method (DDM) to investigate the thermal effects on the permeabil- bining the potential theory for two-dimensional steady flow and
ity of producing an EGS, and McClure and Horne [36] applied the the Runge-Kutta method. The computation procedure is given in
DDM to simulate the heat and mass transfer in discrete fracture Section 6. In the end, the results from the current work are com-
networks. pared with the existing solutions in Section 7 to test the validity
Although the above numerical methods have great advantage in of the proposed model. Finally, more numerical results are pre-
solving complex problems such as irregular geometries or inhomo- sented in Section 8 to find the optimal fracture number and spac-
geneous material, the increased storage memory and thus compu- ing under a given geothermal gradient.
tation time as a result of mesh discretization in FDM, FVM and
FEM, or time convolution in BEM related to time-dependent heat
transfer [37] compromises their uses to some extent in practice, 2. Problem formulation
especially for optimization of the design of an EGS, which requires
a fast algorithm. To the best of the authors’ knowledge, no analyt- The geometry for the present model is shown in Fig. 1. There
ical or semi-analytical solutions are available for the case with exist N (N P 1) parallel horizontal fractures, which provides flow
multiple fractures intercepting dipole wells, and much less work pathways from one injection well (IW) to one production well
has been carried out to investigate semi-analytically the relation- (PW). For the two vertical well system, stress conditions are
ship between the geometrical parameters, including number of assumed to cause hydraulic fractures to grow with a horizontal ori-
fracture, fracture spacing and well separation, and the output per- entation. Indeed, subvertical minimum stress leading to subhori-
formance of an EGS with dipole wells. Therefore, this is the main zontal hydraulic fractures is a relatively common scenario in EGS
motivation of our work for developing a semi-analytical model to [38,39] relative to petroleum reservoirs where the minimum stress
predict the heat extraction from a closed-loop multiple parallel is typically horizontally-directed leading to vertical hydraulic

Fig. 1. Geometry of the model with multiple parallel planar fractures and dipole wells.
 B. Wu et al. / Applied Energy 183 (2016) 1664–1681 1667

fracture growth. Furthermore, multiple, closely spaced, subparallel X

M

hydraulic fractures have been demonstrated in the field provided r  q‘f ¼ Q ‘i dðx  x‘i ; y  y‘i Þ; ð3Þ
i¼1
that each is generated individually from a single, well entry point.
Examples include the Northparkes E48 minethrough experiment where ris the divergence operator and d denotes the Dirac delta,
[40] and the Cadia East minethrough experiment [41,42]. Similar Q ‘i is the flow rate at the ith well ðx‘i ; y‘i Þ in the ‘th fracture with
evidence is reviewed by Bunger et al. [43] along with a model- the sign negative for injection and positive for production. In the
based explanation for conditions under which this geometry is current case, only one set of dipole wells (M = 2) is considered.
expected.
The horizontal fractures are large compared to the well spacing 2.2. Heat transport in closed fractures
and thus taken as infinite, dividing the half space into N + 1 layers.
The whole system (fluid and rock formation) initially is in an equi- According to Cheng et al. [11], the heat transport in the frac-
librium state with the temperature distribution T0 = A0z + B0, tures contains four components, i.e. heat storage, advection, dis-
where A0 is the geothermal gradient, B0 is the ground temperature persion and conduction from the fracture walls, and can be
and z is the depth below the ground surface. The surface tempera- expressed as
ture is assumed to be fixed for all time. When time t > 0, a cold fluid ( )
with a constant injection rate Qin and a constant temperature Tin is @T ‘f q‘f kr @T ‘þ1 @T ‘
þ ‘  rT ‘f  DL r2 T ‘f þ r
 r ¼ 0 at z ¼ z‘0 ;
injected into the IW, and at the same time the fluid is pumped out @t - qw c w -‘ @z @z
at a rate of Qout at the PW. The origin of the coordinate system is set
ð4Þ
to be located at the middle point of the line segment connecting
the two wells on the ground surface. where r2 denote the Laplace operator, T ‘f is the fluid temperature
The geometry of the EGS is defined as follows: the radius and along the ‘th fracture and T ‘r the temperature in the ‘th rock forma-
separation of the wells are rw and 2L, respectively; the depth of tion; qw and cw are the mass density and specific heat, respectively,
the bottom fracture is H and the distance between the ith and jth of the fluid. Additionally, DL is the fluid dispersion coefficient and kr
fractures is denoted by di.j. Without loss of generality, this spacing is the thermal conductivity of the rock formation.
between fractures is taken as a constant throughout the paper. In Actually, the effect of the heat storage and dispersion on heat
addition, the width (aperture) of the ‘th fracture which provides extraction can be ignored without causing significant errors based
the fracture conductivity is denoted by x‘. on the analysis by Cheng et al. [11]. Thus Eq. (3) is rewritten
Some assumptions are made for the present model: ( )
kr @T ‘þ1 @T ‘r
q‘f r T ‘f þ r
 ¼ 0 at z ¼ z‘0 : ð5Þ
(1) The rock is homogenous and impermeable, and the fluid is qw cw @z @z
incompressible;
(2) The physical properties of the rock and fluid are constants Generally there are two types of models describing the heat
independent of temperature change; exchange between the flowing fluid and surrounding rock forma-
(3) Coupled geomechanical mechanisms, such as thermally- tion. The first type, which is most common approach, assumes a
induced strain and/or damage, are not taken into account. temperature continuity across the fracture surfaces, as assumed
That is to say, the fracture apertures are uniform without by [14–16,45,46]. The second type of model characterizes the heat
considering the deformation induced by thermal effect; exchange between fluid and rock by using Newton cooling in terms
(4) As an EGS is generally designed for several tens of years of of heat transfer coefficients (HTCs) whose general formula have
operation, the long-term flow in the fracture can be treated been well developed experimentally in one dimension [47,48],
as steady so that the velocity potential and streamline func- but are not available for two dimensional flow. Zhao [49] devel-
tion method can be applied. oped two analytical solutions based on the above two models,
and showed that the temperature curves from the two models
2.1. Fluid flow in planar fractures almost overlap. Therefore, we adopt the first model here, i.e.

T r‘þ1 ¼ T ‘r ¼ T ‘f at z ¼ z‘0 : ð6Þ

The assumption of steady-state fluid flow is reasonable due to
the fact that fluid flow (pressure diffusion) can reach a steady state
very quickly and the heat exchange addressed in this paper is a 2.3. Heat transport in closed fractures
much longer process. The fluid flow in a plane fracture whose aper-
ture is much smaller than its length is assumed to obey Darcy’s law The heat diffusion equation for each layer of the rock is
[44]
@T ‘r
jr r2 T ‘r ¼ ; ð7Þ
q‘f ¼ r /‘f ; at z ¼ z‘0 ð‘ ¼ 1 . . . NÞ; ð1Þ @t
where the thermal diffusivity jr = kr/qrcr, with qr and cr being the
where r is the gradient operator, /‘f and q‘f ¼ x‘ v ‘f denote the mass density and specific heat capacity, respectively, of the rock
velocity potential and discharge vector, respectively, of the fluid formation.
in the ‘th fracture, in which v ‘f is the fluid velocity vector. In what
follows, vectors and matrices are presented as bold letters. Addi- 2.4. Initial and boundary condition
tionally, z‘0 is the vertical coordinate of ‘th fracture whose sequen-
tial number is counted from the bottom to top The initial and boundary conditions for the whole system are
listed as follows
X
‘1
z10 ¼ H; z‘0 ¼ H  di;iþ1 ð‘ ¼ 2; 3 . . . ; NÞ: ð2Þ T r ¼ T 0 ¼ A0 z þ B0 ; on t ¼ 0;
i¼1 /‘f ¼ /‘0 ; on t ¼ 0;
ð8Þ
The equation of mass continuity for the flow of an incompress- T ‘f ¼ T ‘in ; at injection point ðL; 0; z‘0 Þ;
ible Newtonian fluid in the fracture with M wells is expressed as
follows
T Nþ1
r ¼ B0 ; at z ¼ 0:
1668 B. Wu et al. / Applied Energy 183 (2016) 1664–1681

For the purpose of simplicity, the flow rates at the IW and PW The unknowns in Eq. (12) can be determined by setting some
for each fracture are assumed to be equal, i.e. Q in ¼ Q out ¼ group parameters to be equal to 1 or O(1), i.e.
NQ ‘1 ¼ NQ ‘2 . Therefore the velocity potential /‘f , fluid velocity v‘f
 
and discharge vector q‘f for the fluid are identical for all fractures, G1 ¼ ej2rLt2 ¼ 1; G2 ¼ L/q ¼ 1; G3 ¼ NLQ inq ¼ 2p; G6 ¼ TTr ¼ 1;
T f f f
and thus the superscript ‘ indicating the fracture sequential num- ð13Þ
ber is omitted in subsequent sections. The initial velocity potential G7 ¼ A0 H
T r
¼ g; G8 ¼ L
Lf

¼ 1; / ¼ Q in
2pN
; T r ¼ A0 H þ B0 ;
for flow in a fractured reservoir is denoted as /‘0 where ‘ denotes
the ‘th fracture counted from the bottom. from which we have

3. Scaling analysis
pffiffiffiffiffiffiffi
Q in jr t
T f ¼ T r ¼ A0 H þ B0 ; LT ¼ Lf ¼ L; q ¼ ; e¼ : ð14Þ
2pNL L
To find the controlling evolution parameters, the following
transformation is used Here T r and T f are chosen to be equal to A0H + B0 so that they
cannot be zero.
x ¼ LT ðtÞX T ðtÞ; y ¼ LT ðtÞY T ðtÞ; z ¼ eðtÞLT ðtÞZ T ðtÞ;
In Eq. (12), G11 ¼ g=G9 in which g is a constant approximately
b ‘ ðtÞ; x ¼ Lf X f ; y ¼ Lf Y f ;
z‘0 ¼ eðtÞLT ðtÞ Z equal to 1. The following four controlling parameters are left in
T0 ð9Þ
‘ T ‘r T r ‘ T ‘f T f T in T f /f /0 qf all equations and are redefined as
H ¼ r T r
; H ¼
f T f
; Hin ¼ T f
; U¼ /
; Qf ¼ q
;
pffiffiffiffiffiffiffi
where Li is the length scale, Xi, Yi and Zi are the coordinates (the sub- 2pqr cr NL2 jr t T w  Tf
G4 ¼  K; G5 ¼  Hin ;
script i = f is for fluid flow and i = T for heat flow), e is a small param- qw cw Q in t Tf
eter, H‘in is the dimensionless injection temperature for ‘th fracture,
H 1 d
H‘r and H‘f are the dimensionless temperatures for the ‘th rock layer G9 ¼ pffiffiffiffiffiffiffi  ; G10 ¼ pffiffiffiffiffiffiffi  D;
jr t A jr t
and fracture, U and Qf are the dimensionless fluid velocity potential
and discharge vector, respectively. The variables with superscript ⁄ The physical meanings of these parameters are explained as
are the corresponding scaling parameters. follows:
By substituting Eqs. (9) into (7) the governing equation for heat
conduction in each rock layer is expressed as (1) G4: This is a ratio of heat carried by fluid to that released by
" # rock after a temperature increment of 1 K at a given time t;
t @Tr H‘r jr t @ 2 H‘r @ 2 H‘r 1 @ 2 H‘r (2) G5: This presents the dimensionless inlet temperature,
¼ 2 þ þ ; ð‘ ¼ 1 . . . NÞ ð10Þ
Tr @t LT @X 2T @Y 2T e2 @Z 2T which is taken as a constant;
(3) G9: This provides the length ratio of the heat diffusion zone
which is simplified to be
to the fracture depth;
t @Tr H‘r @ 2 H‘r jr t (4) G10: This provides the length ratio of the heat diffusion zone
 ¼ G1 ; G1 ¼ ; ð11Þ
Tr @t @Z 2T e2 L2T to the fracture spacing.

when the parameter e  1. This means that the heat diffusion can As the injection temperature can be regarded as a constant,
be regarded as one dimensional when the diffusion length scale in there are three evolving parameters for the proposed problem,
z direction is far less than the fracture size. It should be noted that i.e. G9 (1/A), K and D.
the rock temperature is still a function of X and Y although only the So we now have
thermal process in z direction is considered.
The other governing equations together with the boundary and ‘
b T ¼ Z T  G9 ¼ zH ; Z
XT ¼ Xf ; Y T ¼ Y f ; Z b ‘ ¼ Z T  G9 ¼ z0 H ;
initial conditions, i.e. Eqs. (1)–(8), are rewritten as follows eL T T0 eL T

Q f ¼ G2 rUðX f ; Y f Þ; t dD ¼  12 D; t dK ¼  12 K; t dA ¼ 12 G11 ; de t

¼ 12 ;
dt dt dt dt e
r  Q f ðX f ; Y f Þ ¼ G3 ½dðX f  G8 ; Y f Þ  dðX f þ G8 ; Y f Þ;
h ‘þ1 ‘
i
Q f  rH‘f þ G4 @ H r
@Z T
 @@ZHTr ¼ 0; on Z T ¼ Z ‘T0 ð‘ ¼ 1 . . . NÞ; and
H‘f ¼ G5 ; at injection pointðG8 ; 0Þ;
t @Tr Hr D @ Hr K @ Hr A @ Hr Zb @ H‘
H‘f ¼ G6 H‘r on Z ¼ z‘0 =eLT ¼ G9  ð‘  1ÞG10 ; ¼  þ  r
:

Tr @t 2 @D 2 @K 2 @A 2 @Zb
H‘r ¼ Hf =G6 ¼ G7 þ G11 ðZ T  G9 Þ; U‘ ¼ 0; at t ¼ 0;
Therefore, the dimensionless forms of governing equations and
where the parameter groups are listed
boundary conditions for the problems to be solved are reformu-
 kr Tr
G1 ¼ ej2rLt2 ; G2 ¼ L/q ; G3 ¼ Q in
NLf q
; G4 ¼ eTf q qw cw ; G5 ¼ T wTT

0
; lated as follows:
T f f

Tr
G6 ¼ Tf
; G7 ¼ AT0H ; G8 ¼ LL ; G9 ¼ eHLT ; G10 ¼ eLdT ; G11 ¼ A0TeLT ;
r f r (a) Fluid flow problem. The fluid flow that is uncoupled from the
ð12Þ thermal process

with unknown quantities e; LT ; Lf ; / ; q ; T r ; T f :

It should be noticed that the term on the left side of Eq. (11) can r  Q f ¼ 2p½dðX f þ 1; Y f Þ  dðX f  1; Y f Þ;
be expressed in terms of the new evolution parameters ð15Þ
Q f ¼ rU:
t @Tr Hr t dTr
DHr
¼ Hr þt (b) Thermal problem. After solving the fluid flow problem, the
Tr @t Tr dt Dt
X @ Hr dGa flux field can be input as a given quantity that varies
t dTr @ Hr dLT t @ Hr de t
¼ Hr  þt  ZT  ZT : with space but not in time into the thermal problem, defined
Tr dt a @G a dt @Z T dt LT @Z T dt e as
 B. Wu et al. / Applied Energy 183 (2016) 1664–1681 1669

@ H‘r b ‘
@ H‘r 2 ‘
@ H‘r 1
 D2  K2  Z2 @ Hr ¼ @ Hr ; on e  1
þ A2 2
ðX þ coth UÞ þ Y 2 ¼ ; ð20Þ
@b @b
@D @K @A 2
Z Z2 sinh U


‘ @ Hr‘þ1 @ H‘r b¼Zb ‘ ð‘ ¼ 1 . . . NÞ;
Q f  rHf þ K  ¼ 0; on Z for constant U and
@b @b
T0
Z Z
1
H‘f ¼ H‘in ; atð1; 0; Zb ‘T0 Þ; ð16Þ ðY  ctan WÞ2 þ X 2 ¼ 2
; W 2 ½0; p: ð21Þ
sin W
H‘f ¼ H‘f on Zb ¼ ð‘  1ÞD; ð‘ ¼ 1 . . . NÞ;
for constant W, as shown in Fig. 2.
b t ¼ 0;
H‘r ¼ 1 þ A Z; By inversing Eqs. (19) we obtain
H‘f ¼ 1  ADð‘  1Þ; t ¼ 0: sinh U sin W
X¼ ;Y ¼ ; where W 2 ½0; p:
The scaling is therefore used to simplify the governing equa- cosh U cos W cosh U cos W
tions and boundary and initial conditions. The final solution ð22Þ
depends on three dimensionless evolution parameters and also As we take W = p on the straight line connecting the two wells,
on the inlet temperature. However, to find the solution to the the lower signs must be used. Due to symmetry with respect to x
hydrothermal problem it is convenient to use the dimensional axis, only the half plane above the x axis is studied and the map-
forms simplified by fixed characteristic lengths and a single evolu- ping between coordinates (X, Y) and (U, W) is one-to-one. It is
tion parameter, which will be dealt with in the next section. important to mention that W = 0 contains two parts, one going
from the left-most point of the injection well to infinity and the
4. Fluid flow problem other going form the infinity to the production well.
As a pair of conjugate functions, the velocity potential U and
4.1. Dimensionless formulation with fixed characteristic lengths streamline function W satisfy the following identities [44]
@U @W @U @W
To simplify the calculation the following transformation is Q fX ¼  ¼ ; Q fY ¼  ¼ ; ð23Þ
@X @Y @Y @X
utilized
where Q fx and Q fy denotes the dimensionless fluid discharge in the x
z‘0 H
Z ¼ zH
d
; Z ‘0 ¼ d
; s ¼ jdr2t ; H0 ¼ cZ and y direction, respectively. The temperature gradient can be
2 2
ð17Þ expressed in terms of the new variables U and W
a ¼ Hd ; c ¼ A0AHþB
0d
0
; v ¼ Q2pdNkq r Lcw ; v1 ¼ 2pQxinNLd2 j ;
in w r

@ H‘f @ H‘f @ U @ H‘f @ W @ H‘f @ H‘f

where c = ga, H0 is the dimensionless initial temperature and s is ¼ þ ¼ Q fX þ Q fY ;
the dimensionless time. @X @ U @X @ W @X @U @W
ð24Þ
After substituting Eq. (17) into (1)–(8) we obtain the same sim- @ H‘f @ H‘f @ U @ H‘f @ W @ H‘f @ H‘f
plified fluid flow problem which is uncoupled from the thermal ¼ þ ¼ Q fY  Q fX ;
@Y @ U @Y @ W @X @U @W
problem, as given in Eq. (15). As for the thermal problem we have
a simpler form compared with Eq. (16) and thus the advection term in the second of Eq. (18) is simplified to
pffiffiffiffiffi be
@ H‘r @ 2 H‘r jr t
@s
¼ @Z 2
;
 1ð‘ ¼ 1; 2; . . . N þ 1Þ;
on L " #
@ H‘f
h ‘þ1 i @ H‘f @ H‘f
‘ @ Hr @ H‘r ‘ Q f  rH‘f ¼ Q fX Q fX þ Q fY
v1 @ s þ Q f  rHf þ v @Z  @Z ¼ 0; on Z ¼ Z 0 ð‘ ¼ 1; 2; . . . NÞ;
1
@U @W
" #
@ H‘f @ H‘f @ H‘f
þ Q fY Q fY  Q fX ¼ v2 ; ð25Þ
H‘f ¼ H‘in ; at injection pointð1; 0; Z ‘0 Þ; ð18Þ @U @W @U

which is found to be a function of the variable U, and v2 is defined

H‘f ¼ H‘r ¼ Hr‘þ1 on Z ¼ Z ‘0 ;
as follows after combining Eqs. (19) and (22)
H‘r ¼ H0 ¼ cZ; on s ¼ 0; 2
Nþ1 v2 ¼ Q 2fX þ Q 2fY ¼ ðcosh U  cos WÞ ; where W 2 ½0; p: ð26Þ
H r ¼ g; on Z ¼ Z Nþ1
0 ¼ 1=a:
The second equation of Eq. (18), after applying the results of Eq.
It should be noted that the term related to time derivative is
(25) for each fracture, becomes
kept in case some fluid needs much longer travel time from the
‘
" #
IW to the PW. 1 @ Hf @ H‘f @ H‘þ1 @ H‘r
 v2 þv r
 ¼ 0; on Z ¼ Z ‘0 : ð27Þ
v1 @ s @U @Z @Z
4.2. Velocity potential and streamline function
The advantage of the above transformation from Eqs. (19)–(27)
According to the potential flow theory, the dimensionless veloc- is that the original two-dimensional PDE is converted into a one-
ity potential U and streamline function W, based on Eq. (15), can dimensional PDE with respect to the variable U, and then it
be obtained by using the superposition of the solution for a single becomes an ODE after applying the Laplace transformation.
source or sink. The solutions for the Laplace equation of the single The flow in the fracture can be regarded as a series of fluid par-
source problem are given as [44] ticles moving along the streamlines, as shown in Fig. 2. As men-
" # tioned above, the coefficient of heat storage term is used to
1 ðX  1Þ2 þ Y 2 2Y calculate the travel time of fluid particles. By combining it with
U¼ ln ; tan W ¼  ; W 2 ½0; p:
2 ðX þ 1Þ2 þ Y 2 1  X2  Y 2 Eq. (26), the time for the injected fluid to reach a given point (U,
ð19Þ W) along the streamline W is expressed as

Therefore, the 2-D fracture plane can be reconstructed by the

Z þ1
1 df
velocity potential and the streamline function, in terms of a series sh ðU; WÞ ¼¼ gðU; WÞ; gðU; WÞ ¼ ; ð28Þ
v1 U ðcosh f  cos WÞ
2
of orthogonal circles, i.e.
1670 B. Wu et al. / Applied Energy 183 (2016) 1664–1681

Fig. 2. New orthogonal coordinate system for the fracture plane in terms of variables U and W. The red dashed curves represented by Eq. (20) denote the points where the
velocity potential U are equal, and the blue solid ones represented by Eq. (21) denote the streamlines with the same value W.

where g(U, W) can be integrated analytically, as provided in the where s is a complex number and the cap above a certain variable
appendix, and is obtained as follows denotes the Laplace transform of the corresponding variable. For

(   
Ucos W þ 2 cot W arctanðcot 2 Þ  arctan tanh 2 cot 2
1  coshsinh U W U W 1
sin2 W
; onj cos Wj–1;
g¼  cos W 
1
3
1 þ coshsinh
Ucos W cosh Ucos W  1 ; onj cos Wj ¼ 1:
U

From the above equation it is apparent that the travel time for a example, H ^ ‘ are the Laplace transforms of H‘ and H‘ ,
^ ‘ and H
r f r f
fluid particle to reach the PW is determined by setting U to 1. If respectively; A‘ and B‘ are the undetermined coefficients related
s < sh(1, W). The consequence is that a cool fluid particle along to the ‘th rock layer.
this streamline does not reach the PW and thus the temperature Substitution of Eq. (29) into both the second equation in Eqs.
at the intercepting point of this streamline with the PW does not (18) and (27) leads to
change.
One interesting parameter in the EGS design is the thermal b‘ ðA‘þ1  A‘ Þ ¼ ðB‘þ1  B‘ Þ;
breakthrough time (TBT), which is defined as the shortest time ðb‘ A‘ þB‘ Þs
  pffiffi ð30Þ
v  v2 b‘ @A
@U
‘
þ @B
@U
‘
þ 2v sb‘ ðA‘þ1  A‘ Þ ¼ 0;
for a fluid particle to reach the PW among all the streamlines 1

[15]. In the current case, the TBT corresponds to the fluid particle pffiffi
where b‘ = expð2 sZ ‘0 Þ:
flowing along the streamline along W = p, since the straight line
Application of the boundary conditions at the ground surface
connecting the dipole wells is the shortest, and the TBT is found
to be equal to 2/(3v1). (Z ¼ Z Nþ1
0 ¼ 1=a), Z ? +1 and at the injection point gives

A1 ¼ 0; BNþ1 ¼ ANþ1 bNþ1 ;

5. Thermal problem pffiffiffiffiffi ð31Þ
b‘ A‘ þ B‘ ¼ b‘ ðH‘in  cZ ‘0 Þ=s on U ! þ1:
5.1. Laplace transform Eq. (30) and the boundary condition (31) can be written in a
matrix form
By applying the Laplace transform to Eq. (18) with the initial
conditions, we obtain the transformed temperature for the ‘th CA ¼ DB;
(counted from the bottom to top) rock layer and the fluid in the pffiffi  
v1 ½FA þ B þ 2v sEA  v2 F ddAU þ ddBU ¼ 0;
s
ð32Þ
‘th fracture
pffi pffi FA þ B ¼ WV; when U ! þ1;
^ ‘ ¼ A‘ ðU; W; sÞe
H sZ
þ B‘ ðU; W; sÞe sZ þ csZ ;
r
pffi ‘ pffi ‘ cZ ‘ ð29Þ where the unknown vectors A and B, the known vector V, and the
^‘ ¼ H
H ^ ‘j ‘ ¼ A‘ ðU; W; sÞe sZ0 þ B‘ ðU; W; sÞe sZ0 þ s 0 ;
f r Z¼Z 0 known matrices C, D, E, F, W are defined as follows
 B. Wu et al. / Applied Energy 183 (2016) 1664–1681 1671

A ¼ ½ANþ1 ; AN ; AN1 ; . . . ; A4 ; A3 ; A2 T ; B ¼ ½BN ; BN1 ; BN2 ; . . . ; B3 ; B2 ; B1 T ;

2 3
bN  bNþ1 bN 2 3
6 7 1
6 . 7
6 bN1 . . 7 6 1 1
6
7
7
6 7 6 7
6 . . 7 6 1 1 7
6 . . . . 7 6 7
6 7 6 . . 7
C¼6
6 .. .. 7
7 ; D ¼ 6 . . . . 7;
6 . . 7 6 7
6 7 6 1 1 7
6 7 6 7
6 b 3 b 3 7 6 7
6 7 4 1 1 5
4 b2 b2 5
1 1
b1
2 3 2 3
bN bN 0 bN
6 7 6 0 bN1 7
6 bN1 bN1 7 6 7
6 7 6 7
6 bN2 bN2 7 6 . . 7
6 7 6 0 . 7
6 . . 7 6 7
.. .. 6 . 7
E¼6
6
7; F ¼ 6
7 . . 7;
6 7 6 7
6 . . . b 7 6 . 7
6 7 6 0 .. 7
6 3
7 6 7
4 b2 b2 5 6 7
4 0 b2 5
b1 0
2 pffiffiffiffiffiffi 3 2 3
bN HNin  cZ N0
6 pffiffiffiffiffiffiffiffiffiffi 7 6
6 7 N1 7
6 HN1 7
6 in  cZ 0 7
6 bN1 7
6 .. 7 6 N2 N2 7
6 7 6 Hin  cZ 0 7
6 . 7 6 71
6 7
W¼6 .. 7; V ¼ 66 .. 7
7 ;
6 . 7 6 . 7s
6 pffiffiffiffiffi 7 6 7
6 7 6 H3in  cZ 30 7
6 b3 7 6 7
6 p ffiffiffiffiffi 7 6 7
6 7 2
4 Hin  cZ 0 5
2
4 b2 5
pffiffiffiffiffi 1 1
b1 Hin  cZ 0

where the empty spaces contain zero elements. 5.2. Fracture-independent model
By using Eq. (26) and replacing B with A, Eq. (32) becomes
h pffiffi i This is a special case where heat diffusion is very slow in the
dA
dU
¼ ðcosh Ucos
1
W Þ2
2v sM þ vs I A; rock formation. The diffusion velocity can be roughly calculated.
1

A ¼ Q; when U ! 1; ð33Þ According to JiJi [50], for a semi-infinite domain with a homoge-
nous initial temperature or a time-dependent surface heat flux
B ¼ M1 A;
imposed at the boundary at t > 0, the penetration depth of the ther-
where mal layer is in the order of (6jrt)1/2. If we take the properties of
granite, for example, then kr = 3.0 W/(m K), qr = 2700 kg/m3 and
M ¼ ðF þ M1 Þ1 E; M1 ¼ D1 C; Q ¼ ðF þ M1 Þ1 WV: cr = 790 J/(kg K), and the thermal diffusivity jr = 1.406 106 m2/
s. After 30 days, 1 year, 10 years and 30 years, the depths of the
Therefore, the original problem is converted to the problem of
thermal layer are 4.67 m, 16.31 m, 51.59 m and 89.35 m, respec-
solving a system of first order ODEs with respect to U under the
tively. This means that, at early time or when the fracture spacing
boundary conditions at U ? +1.
is large enough, the thermal penetration layers surrounding the
For the purpose of simplifying the programing, the following
fractures are much smaller than the fracture spacing. In this case,
transformation is used so that U ? ±1 is avoided
the temperatures for the rocks surrounding each fracture can be
q ¼ eU ; ddUq ¼  q1 ; q 2 ½0; þ1Þ; calculated independently.
ð34Þ If the thermal interaction between fractures is ignored, the ana-
Aðq; W; sÞ ¼ AðU; W; sÞ; Bðq; W; sÞ ¼ BðU; W; sÞ; lytical solutions to Eqs. (30) and (31) for the temperatures of the
where q is used as the new variable replacing U. The points at U ? fluid in each fracture and the rock formation in each layer can be
+1 and U ? 1 correspond to those at q = 0 and q ? +1, respec- obtained in a similar way as that for the case with a single fracture,
tively. It has to be noted that the range of U along the streamline as shown in the appendix, leading to
p ffis
W = 0 is [0, +1), and thus q 2 [0, 1] for this case. gðU;WÞ  psffi z pffi 
2v
^ ‘þ1 ¼ H‘ ð1b
H in
e ‘þ1 Þ e  b‘þ1 e sz
þ csZ ;
Eq. (33) is re-written in terms of variable q as follows r sð1b‘þ1 Þ
p ffis pffi ð36Þ
gðU;WÞ pffisZ
2v 2v s
^‘ ¼ H‘in ð1b ‘ ‘
^ ‘ ¼ Hin eð1b‘þ1 ÞgðU;WÞ þ cZ0 ;
dA=dq ¼ f ðqÞKA ¼ Uðq; AÞ; Hr s
e ‘þ1 Þ e þ csZ ; H f s s

A ¼ Q ; when q ¼ 0; ð35Þ where the function g(U, W) is defined above. From the above equa-
M1 A ¼ B; tion, the outlet temperature is directly obtained.
If the effect of the geothermal gradient is not considered and the
where the function f(q) and the matrix K are defined depth of the bottom fracture H is large enough, the solution can be
2 pffiffi simplified to be a semi-infinite one provided by Schulz [15] except
f ðqÞ ¼ 4q=ðq2 þ 1  2q cos WÞ ; K ¼ 2v sM: that the storage effect is not considered by Schulz [15], as follows
1672 B. Wu et al. / Applied Energy 183 (2016) 1664–1681

H1in pffi pffi tions, the point representing the well should not be considered
^1 ¼
H e2e sgðU;WÞ sZ
ðZ P 0Þ; ð37Þ
r
s as part of the flow domain because the Laplace’s equation does
not apply there. We only need to consider the temperature at the
for which the solution in the time domain is
points on a small circle around the PW, where the value of q is
 
2vgðU; WÞ þ Z finite. Then the outlet temperature of the ‘th fracture is defined
H1r ¼ H1in erfc pffiffiffi ðZ P 0Þ: ð38Þ as the averaged value of the temperatures of the points on this cir-
2 s
cle, i.e.
Z 2p Z 2p X
M X
M
5.3. Multi-fracture model
T ‘out ¼ T ‘f q‘f dh= q‘f dh T ‘f q‘f = q‘f :
0 0 j¼1 j¼1
When the fracture spacing is small or the time is large, the ther-
mal interaction effect cannot be ignored. In this case, the coupled The steps to calculate the outlet temperature are listed as
problem has to be solved. Although Eq. (35) is a linear system of follows:
the first order ODEs, when the number of fractures is larger than
one, it is impossible to obtain the analytical solutions because of 1. For any given time t = t0;
the variable coefficients, tridiagonal but non-symmetric matrix M 2. Start with the first point of first streamline, q = 0 and W = p;
and the large matrix dimensions. Therefore, we have to rely on a 3. Obtain A and B at the injection point q = 0 (dA=dq ¼ 0 and
numerical approach. In the current work, the fourth order Runge- dB=dq ¼ 0);
Kutta method [51] is applied 4. Move the point to the new position q = q + Dq with the same
k1 ¼ hUðqn ; An Þ; W;
5. Solve for A and B at the current position (q, W) by using the
k2 ¼ hUðqn þ h=2; An þ k1 =2Þ;
Runge-Kutta method, and thus obtain A and B at the point (U,
k3 ¼ hUðqn þ h=2; An þ k2 =2Þ; W);
k4 ¼ hUðqn þ h; An þ k3 Þ; 6. Obtain the numerical value in the time domain for the rock
5 temperature by using the inverse Laplace transformation [52];
Anþ1 ¼ An þ k1 =6 þ k2 =3 þ k3 =3 þ k4 =6 þ Oðh Þ;
7. Repeat steps 4–6 and obtain the temperature for the points
where h is the spatial step with respect to q. along the streamline with the same W until the PW is reached;
When A (q, W) is solved, B (q, W) is obtained according to Eq. 8. Start with the first point of a new streamline, follow the steps
(35). Then the vectors A and B at the point (U, W) are known. After from 3 to 7 and obtain the temperature information for all the
substituting A and B into Eq. (29), we can calculate the tempera- points along this new streamline. The calculation is repeated
ture for the rock formation and thus for the fluid in the fracture until all the streamlines are visited so that the temperature field
using a proper inverse method for Laplace transformation. in the half plane is obtained for the current time t = t0.

Start a new time t and obtain the rock and fluid temperatures by
6. Computational procedure
repeating steps 2–8.

Because the fluid flow is symmetrical to the x-axis, only the half
plane above the x-axis is included in the calculation, as shown in 7. Verification and sensitivity study
Fig. 3. Firstly, the semi-circle of the IW is evenly divided by NC sub-
divisions. Next, the value of the streamline function along each of 7.1. Model verification
the NC + 1 streamlines, Wk (k = 1, 2. . .NC + 1), is obtained. All the
streamlines will flow into the PW except the one with W = 0, which In this subsection, the results from three approaches, i.e. the
flows away from the IW to infinity. As mentioned in the above sec- fourth order R-K method (RK) based on Eq. (35), analytical solution
tions, the range of the new variable is q 2 [0, +1]. In the calcula- (AS) based on Eqs. (37) or (40) and the exact solution (ES) in the

Fig. 3. Computational procedure.

 B. Wu et al. / Applied Energy 183 (2016) 1664–1681 1673

time domain given by Eq. (38), are compared to each other for the move towards to the PW with time increasing; and (2) the results
single fracture case. The Stehfest [52] algorithm, which is used for from the three methods are in excellent agreement, indicating that
obtaining the inverse Laplace transformation in the present calcu- the numerical results from the RK method are correct.
lation due to its ease of use and high accuracy, takes the form
7.2. Sensitivity study
ln 2 X
N
ln 2
f ðtÞ ¼ cn f n ;
t n¼1 t
Because the outlet temperature for each fracture is averaged for
where the coefficients cn is defined as the points on a small circle close around the PW, a sensitivity study
of the outlet temperature on the size of the circle and the number
X
minðn;N=2Þ N=2
k ð2kÞ! of the streamline flowing into the PW is necessary. Here, we car-
cn ¼ ð1ÞnþN=2 :
k¼ðnþ1Þ=2
ðN=2  kÞ!k!ðk  1Þ!ðn  kÞ!ð2k  nÞ! ried out the mesh sensitivity study for the single fracture case
when A0 = 0.
The optimal number N = 6 in this algorithm is found to give Fig. 5(a) shows the output temperature for different radii r0
accurate results for the present calculation although it ranges from (ranging from 0.05 m to 0.3 m) of the circle and Fig. 5(b) for differ-
6 to 20 according to Cheng et al. [53]. The parameters used for the ent numbers of streamlines (ranging from 51 to 251). For both
subsequent calculations are listed in Table 1, unless elsewhere cases, the physical and geometrical parameters are kept
specified. unchanged. In Fig. 5(a) the curves for the output temperature coin-
Fig. 4(a) and (b) displays the fluid temperature along the cide indicating that the average temperature is not sensitive to the
straight streamline connecting the two wells for two cases, i.e. zero size (at least in the range of 0.05–0.30 m) of the circle. r0 = 0.1 m is
(A0 = 0) and nonzero (A0 – 0) geothermal gradients, respectively. chosen for the subsequent calculations.
The initial temperature (T0 = 238.5 °C) is the same everywhere for The outlet temperature for different numbers of streamlines is
the first case, while this initial temperature is applied to the depth provided in Fig. 5(b). In this case, NC of a total NC + 1 streamlines
of H = 4500 m for the second case. The dimensionless time diverging from the IW and converging into the PW are used to cal-
s = 0.018, 1.77 and 53.3 correspond to the physical 0.01, 1.0 and culate the temperature at the points on the circle. The streamline
30 years, respectively. In Fig. 4(a) the results obtained from the W = 0 from infinity does not contribute to the temperature change
above three approaches, are plotted, while in Fig. 4(b) only the of the PW because the fluid temperature is always equal to the ini-
results from the RK and AS methods are given due to unavailable tial temperature. From Fig. 5(b) it is apparent that the difference in
exact solutions in the time domain. It can be seen from these the production temperatures when NC = 50, 150 and 250 is so
two figures that: (1) the temperature on this streamline shows a small that it can be neglected. Hence, NC = 150 is chosen for the
very smooth increasing trend from IW to PW and the cooling fronts subsequent calculations.

Table 1
Parameters for the calculations.

Parameter Value Parameter Value

Wellbore radius, rw (m) 0.1 Fracture aperture, w (m) 0.001
Well separation, re (m) 500 Depth of bottom fracture, H (m) 4500
Granite thermal conductivity, ks (W m1 K1) 3.0 Injection temperature (°C) 90
Granite mass density, qr (kg/m3) 2700 Geothermal gradient, A0 (°C/m) 0.047
Granite specific heat, cr (J kg K1) 790 Surface temperature, B0 (°C) 27
Fluid thermal conductivity, kw (W m1 K1) 0.6 Total flow rate, Q (m3/s) 0.150
Fluid mass density, qw (kg/m3) 1000 Fracture number, N Varies
Fluid specific heat, cw (J kg K1) 4200 Fracture spacing, d (m) Varies

0.0 0.0
RK, τ=0.018 RK, τ=0.018
AS, τ=0.018 AS, τ=0.018
-0.2 ES, τ=0.018 -0.2 RK, τ=1.77
RK, τ=1.77 AS, τ=1.77
Temperature (Θ f )

Temperature (Θ f )

AS, τ=1.77 RK, τ=53.2

ES, τ=1.77 AS, τ=53.2
-0.4 -0.4
RK, τ=53.2
AS, τ=53.2
ES, τ=53.2
-0.6 -0.6

-0.8 -0.8

injection Production injection Production

-1.0 -1.0
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
Distance X Distance X
(a) (b)
Fig. 4. Comparison of the temperature along the line connecting the two wells for single fracture case: (a) A0 = 0, B0 = 238.5 °C, (b) A0 = 0.047 K/m, B0 = 27 °C and H0 = 4500 m.
1674 B. Wu et al. / Applied Energy 183 (2016) 1664–1681

0.0 0.0
r0 =0.05 m, RK NC= 50, RK
r0 =0.05 m, ES NC= 50, ES
-0.1 r0 =0.15 m, RK -0.1 NC=150, RK
NC=150, ES

Output temp. (Θout )

Output temp. (Θout )

r0 =0.15 m, ES NC=250, RK
r0 =0.30 m, RK NC=250, ES
-0.2 r0 =0.30 m, ES -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
0 5 10 15 20 25 30 5 10 15 20 25 30
Time (years) Time (years)

(a) (b)
Fig. 5. Comparison of output temperature for the single fracture case: (a) for different sizes of artificial circles, (b) and for different number of streamlines.

8. Numerical results 8.2. Temperature in surrounding rock formations

8.1. Temperature of the fluid in fractures In order to show the effect of the fracture spacing on the ther-
mal behavior in the multiple fracture system, the profiles of rock
Fig. 6(a) gives a general view for the temperature variation of temperatures when the number of fractures is N = 6 and for differ-
the fluid in a multiple fracture system, and Fig. 6(b) provides a ent fracture spacings are plotted in Fig. 7. Because the rock temper-
detailed view around the PW. As the steady-state fluid flow path ature field is three-dimensional, we consider the contours of the
is fixed, only the first (bottom) fracture is studied. The number of temperature change on a vertical section containing the IW and
fractures is N = 3, fracture spacing d = 60 m and total injection rate the PW.
Qin = 0.45 m3/s. By using the symmetry of the model geometry Fig. 7(a) shows the results when the fracture spacing is
with respect to x axis, only the half plane above the x axis is plot- d = 10 m, while the well distance is 1000 m. As mentioned above,
ted. As the total injection rate is assumed to be evenly shared by all the thermal layer can be roughly calculated by using (6jt)1/2, and
fractures, the inject rate for one individual fracture is 0.15 m3/s in is about 5.15 m at time t = 0.1 years in this case. That is to say,
this case. when the EGS operation time is larger than this value, the thermal
From Fig. 6 we see the propagation of the thermal front on the interaction among fractures must be taken into account. With fur-
fracture plane. Some factors, such as the injection temperature, ther increase of time, heat supply to each fracture tends to be
pumping rate and temperature of the surrounding formation, con- homogeneous and the initial multiple thermal affected zones col-
trol the thermal behavior. In an EGS, the injection temperature of lapse to one, leading to a low temperature region in the reservoir
the fluid can be considered as constant. The temperature of sur- zone (at the depth between 4450 m and 4500 m). The low temper-
rounding formation determines the amount of heat remaining in ature zone becomes larger and larger until it passes the PW, as
the rock to be removed by the flowing fluid. The pumping rate is shown in Fig. 7(a), when t = 30 years. Eventually, the heat extrac-
one of most important parameters to determine how much and tion from the production well becomes less efficient and the life-
how fast the heat can be produced from the EGS in a given opera- time of the EGS ends.
tional time. The thermal breakthrough time (TBT), which is related The EGS’s reservoir’s useful life is extended when the fracture
to the pumping rate and is defined as the smallest time for the spacing is increased to 100 m, as shown in Fig. 7(b) where other
thermal fronts of all the streamlines to reach the PW, is of interest parameters are kept the same as Fig. 7(a). Firstly, the thermal
and importance. As shown in Eq. (28) the function g(1, W0)/v1 affected zones surrounding each fracture evolves independently
denotes the time for the thermal front to reach the PW along a until a much longer time. Secondly, because more heat is absorbed
streamline function W0. In the present case, the TBT g(1, p)/ from the surrounding rock, the fluid can maintain a much higher
v1 = 2.73 106 corresponds to t = 6981 s, the time required for temperature later in the production life. Thirdly, the interaction
the fluid to flow along the straight streamline connecting the between the thermal affected zones of neighbouring fractures is
dipole wells. Although some cooler fluid from the IW has reached weak so that they never collapse into one during the assigned
the PW by t = 0.1 years, the induced temperature change around EGS operation lifetime (30 years). Finally, even after 30 years, the
the PW is trivial, as shown in Fig. 6. This is due to the fact that rock around the PW still has a high temperature, indicating suffi-
the fluid, although with a much lower initial temperature, is cient heat supply.
heated to a higher temperature through transfer of heat from the
rocks during the period it takes to flow towards the PW. However, 8.3. Production temperature under different number of fractures N and
with increasing time, the zone of rock affected by cooling becomes different fracture spacing d
larger and larger as shown in Fig. 6, leading to a cooler fluid reach-
ing the PW seen, for example, when comparing the temperature for As can be seen in Eq. (16), the heat transfer process is mainly
production times at 0.5 years and at 30 years. controlled by three parameters, i.e. A, D and K which can be
 B. Wu et al. / Applied Energy 183 (2016) 1664–1681 1675

Fig. 6. Fluid temperature in the first fracture for multiple fracture EGS with N = 3, d = 60 m, Qin = 0.45 m3/s. (a) general view for the region around the dipole wells, and (b)
magnified view at the PW. The dimensionless coordinates are defined as X = x/L and Y = y/L.

regarded as the dimensionless geothermal gradient, dimensionless D = d/(jrt)1/2, reflects the potential for thermal interaction between
pffiffiffi
fracture spacing, and dimensionless thermal conductivity of the fractures. In limiting cases, when d/d < 1/2 or D > 2 6, each frac-
fracture interface, respectively. Calculation of the PT requires ture works independently. In contrast, when D ? 0 (i.e. at suffi-
another parameter, i.e. number of fracture N. ciently small spacing or large time), the EGS will behave like a
Although the parameter A = (jrt)1/2/H is very small throughout single fracture from the perspective of the thermal problem. Addi-
our regime of interest as the diffusion length d = (6jrt)1/2  H, tionally, the system evolves from an early time regime in which
the contribution that comes via the initial conditions cannot be advective flow can be neglected to a large time regime in which
neglected. Additionally, the parameter comprised of the ratio of the temperature gradient in the rock vanishes (i.e. a uniform tem-
the spacing to the characteristic length of thermal diffusion, perature is attained) and therefore heat transfer between the rock
1676 B. Wu et al. / Applied Energy 183 (2016) 1664–1681

Fig. 7. Contours of temperature change in the vertical cross section connecting the dipole wells for small and large fracture spacings: (a) N = 6, d = 10 m; (b) N = 6, d = 100 m.

and fluid is negligible. According to Eq. (16), the parameter 120 m and when N is 2, 6 and 10, as shown in Fig. 8(a), (b) and (c),
pffiffiffiffiffiffiffi
K ¼ ð2pkr NL2 Þ=ðqw cw Q in jr t Þ embodies this aspect of the evolu- respectively. In all three cases, other parameters, including the
tion of the system. Because heat transfer to the fluid is poor both total injection rate, are kept the same.
in the cases of advective heat transfer and vanishing thermal gra- When the number of fractures is small (for example N = 2 in
dient, an EGS must operate at intermediate values of this quantity. Fig. 8(a)), even large fracture spacing does not increase the PT sig-
The above three parameters {A, D, and K} are dependent upon nificantly, not to mention small fracture spacing. The change of PT
the number of fractures (N) and fracture spacing (d). Hence, the fol- is very small even when d increases from 80 to 120 m.
lowing parametric study focuses on the dependence of the solution As indicated in Eq. (16), given the same total injection rate, the
for the production temperature (PT) on N and d, leading to a con- effect of the number of fractures, N, on the PT lies in two aspects:
clusion that judicious choice of these quantities can optimize the (1) ensuring suitable heat transfer during the life of the EGS, which
economical lifetime of an EGS well system. is related to the parameter K, and (2) balancing the fact that greater
spacing both promotes sustained PT by increasing the volume of
8.4. Varying d and fixing N rock from which heat can be extracted while also pushing the array
of fractures to shallower, cooler depths thereby leading to overall
The efficiency of multiple fractures in extracting heat from the lower PT due to averaging of the production temperatures from
rock is reflected by the production temperature (PT) of the EGS. all the fractures. More specifically, if N is fixed to be small, as there
Fig. 8 gives the PTs for different values of d ranging from 5 m to are only a maximum difference of 5.64 °C in original temperatures
 B. Wu et al. / Applied Energy 183 (2016) 1664–1681 1677

0.0 0.0

(a) d = 5m
d = 10m (a) N =1
N =2
d = 20m N =3
-0.1 d = 40m -0.1 N =5
Prodution temp. Θ

Prodution temp. Θ
d = 60m N =7
d = 80m N =9
d =100m N =11
d =120m N =13
-0.2 -0.2

-0.3 -0.3

-0.4 -0.4
5 10 15 20 25 30 5 10 15 20 25 30
Time (years) Time (years)

0.0 0.0

(b) (b)
-0.1 -0.1
Prodution temp. Θ

-0.2 Prodution temp. Θ -0.2

d = 5m N =1
d = 10m N =2
d = 20m N =3
d = 40m N =5
-0.3 d = 60m -0.3 N =7
d = 80m N =9
d =100m N =11
d =120m N =13
-0.4 -0.4
5 10 15 20 25 30 5 10 15 20 25 30
Time (years) Time (years)

0.0 0.0

(c) (c)
-0.1 -0.1
Prodution temp. Θ

Prodution temp. Θ

-0.2 -0.2
d = 5m N =1
d = 10m N =2
d = 20m N =3
d = 40m N =5
-0.3 d = 60m -0.3 N =7
d = 80m N =9
d =100m N =11
d =120m N =13
-0.4 -0.4
5 10 15 20 25 30 5 10 15 20 25 30
Time (years) Time (years)

Fig. 8. Production temperature change of EGS normalized by A0H + B0 for different Fig. 9. Production temperature change of EGS normalized by A0H + B0 for different
fracture spacings. (a) N = 2, (b) N = 6, and (c) N = 10. numbers of fractures. (a) d = 5 m, (b) d = 50 m and (c) d = 100 m.

of neighboring fractures (maximum spacing 120 m in the current can be improved to some extent by increasing the number of frac-
case), the PT is fracture spacing dominated in terms of D. In this tures, N. When N becomes larger, in addition to the increased con-
case, the larger fracture spacing d, the less interaction between tacting area for heat exchange between the flowing fluid and the
fractures and thus the better output performance. When the frac- surrounding rock, the larger dimensionless thermal conductivity,
ture spacing increases, the thermal interaction becomes weaker K, of fracture interface, indicating a quicker heat exchange rate,
and weaker until the PT will not significantly change even as the also contributes to the improving the PT. Eventually the fracture
fracture spacing continues to increase. spacing D and the number of fracture N will result in the geother-
Numerical experiments illustrate these behaviors. From Fig. 8 mal gradient A becoming important, but prior to this the PT is frac-
(b) and (c) we observe firstly that the production performance ture number dominated. Under this condition, larger N leads to
1678 B. Wu et al. / Applied Energy 183 (2016) 1664–1681

larger PT. Secondly, the PT will decrease when the fracture spacing 8.6. Effect of the geothermal gradients on the PTs
is larger than a threshold value. This is a process involving the
competition between thermal interaction related to dimensionless Fig. 10 displays the evolution of production temperature change
fracture spacing D and contribution of initial conditions related to in a multiple fracture EGS with the number of fractures N = 10 and
the dimensionless geothermal gradient A. When D becomes large fracture spacing d = 60 m for zero and nonzero geothermal gradi-
enough, the positive contribution to the PT as a result of weaker ents A0. In order to emphasize the geothermal gradient effect, the
thermal interaction between fractures is compensated by a nega- initial temperature at a depth of 4500 m in the case of nonzero
tive contribution associated with the initial conditions, i.e. the fact A0 (B0 = 27 °C) is equal to the homogenous initial temperature in
that the more widely-spaced array extends to shallower, and the case with A0 = 0 (B0 = 238.5 °C). If there is only one fracture in
therefore initially cooler, depths. In other words, the shallower the EGS, for example in the model used by Cheng et al. [11], Schulz
rock layers, which are at a lower temperature, dominate over the [15] and Ghassemi et al. [33], the assumption of a zero geothermal
deep layers in the final PT when the fracture spacing and number gradient does not lead to significant error in the PT prediction. This
of fractures are large. This threshold value of D where further is due to the fact that, as mentioned above, the maximum temper-
increase in fracture spacing becomes counterproductive is smaller ature difference in the region where the thermal layer penetrates
for larger numbers of fractures. after 30 years is around 3.3 K due to the slow heat diffusion. How-
ever, for the multiple fracture case, as shown in Fig. 10, the differ-
ence in the production temperatures predicted from both scenarios
8.5. Varying N and fixing d
is roughly 12 K.

To further clarify the effect of N and d on the production perfor-

mance, the PTs for different numbers of fractures are demonstrated 8.7. N-d curve for the PTs
in Fig. 9. Here the fracture spacing is 5, 50 and 100 m in Fig. 9(a),
(b) and (c), respectively. In the above subsections, the effects of the number of fractures,
As discussed above, the thermal behavior in the rock layers of a fracture spacing and geothermal gradient on the PTs of an EGS are
multiple fracture EGS is similar to that of a single fracture EGS studied individually. Fig. 11 presents the PT variations in terms of
when the spacing is small. Fig. 9(a) confirms this conclusion in both N and d. It can be seen that the cases with small N and small d
terms of PTs by indicating that there is no large change in the PT (lower left corner in the figure) display the worst performance for
when the number of fractures varies from 1 to 13. As calculated the EGS. The zone with relatively high temperature (denoted by
above, the heat diffusion length after 30 years is roughly equal to red colour) occupies most of the central part before an operation
d = (6jrt)1/2 = 89.35 m, while the distance between the top and time of t = 16 years, and then moves towards the upper middle
bottom fractures is only 60 m. In this case, the influential zone of part. For the given parameters, d = 50 m produces the highest PT
all fractures overlaps after 30 years, making the EGS work like a when t = 30 years as shown in Fig. 11, where the PT still increases
single-fracture. if the number of fractures N is increased to above 13. However, the
In addition, the curves of the PT changes in Fig. 9(b) and (c) PT in the region enclosed by N 2 [6, 13] and d 2 [50, 90] is high
show similar trends as portrayed in Fig. 8. On the one hand, by enough to provide sustainable heat supply. Therefore, these num-
increasing the fracture spacing to 50 m and 100 m, as shown in bers provide the good choices for the design of an EGS with an
Fig. 9(b) and (c), respectively, the PT can be increased to some operation lifespan of 30 years.
extent depending on the values of N and d. On the other hand,
there also exists a threshold value for the number of fracture N, 9. Conclusion
above which the PT will decrease. Similar to the case in Fig. 8, it
is shown that larger fracture spacing leads to smaller threshold In this paper, by using velocity potential and streamline func-
values of N for which additional fractures become counterproduc- tions to replace real coordinates, the two-dimensional equation
tive. Therefore, the geothermal gradient is an important factor in for fluid flow in planar fractures is reduced to a one-dimensional
controlling the PT of an EGS and ignorance of its effects on the ther- equation and a semi-analytical model is developed to predict the
mal behavior of an EGS may lead to less efficient designs. heat extraction from a multiple parallel fracture closed-loop EGS
using dipole wells. The solution is applied to study the effect of
the number of fracture and fracture spacing on the production
0.00
temperature of the EGS. Based on the numerical results, some con-
A0=0, B0=238.5 oC clusions are obtained.
-0.05 A0=0.047 oC/m, B0=27 oC

(1) The semi-analytical solutions provide accurate results, and

Production temp. Θ

-0.10 serve as a benchmark for numerical methods;

(2) The reservoir’s lifetime can be significantly increased when
-0.15 multiple fractures are created, compared to a single fracture
reservoir;
(3) The spacing, d, and number, N, of fractures for an EGS with
-0.20
multiple fractures are two important factors controlling
the production temperature. Multiple fractures can be used
-0.25 to maintain the production temperature of an EGS by
increasing the contacting surface area between the flowing
-0.30 fluid and the rock formation. However, if the fracture spac-
5 10 15 20 25 30 ing is too small or the production time sufficiently long the
Time (years) multiple fracture system works like a single fracture;
(4) The geothermal gradient is an determining parameter on the
Fig. 10. Effect of the geothermal gradient on the production temperature change
production performance including the optimal number and
normalized by A0H + B0 when the number of fracture is N = 10 and fracture spacing
is d = 60 m. spacing of fractures;
 B. Wu et al. / Applied Energy 183 (2016) 1664–1681 1679

Fig. 11. Contours for production temperature change of EGS normalized by A0H + B0 for different fracture spacings d and different numbers of fractures N.
1680 B. Wu et al. / Applied Energy 183 (2016) 1664–1681

(5) The semi-analytical solutions can also be used for modelling References
shallow geothermal reservoirs;
(6) The proposed model is effective and efficient in computa- [1] de Moel M, Bach PM, Bouazza A, Singh RM, Sun JO. Technological advances and
applications of geothermal energy pile foundations and their feasibility in
tions. It provides a useful tool for design and optimization Australia. Renew Sustain Energy Rev 2010;14:2683–96.
of an enhanced geothermal system. [2] Ally MR, Munk JD, Baxter VD, Gehl AC. Exergy analysis of a two-stage ground
source heat pump with a vertical bore for residential space conditioning under
simulated occupancy. Appl Energy 2015;155:502–14.
[3] Priarone A, Fossa M. Modelling the ground volume for numerically generating
Acknowledgements single borehole heat exchanger response factors according to the cylindrical
source approach. Geothermics 2015;58:32–8.
[4] Kuzmic N, Law YLE, Dworkin SB. Numerical heat transfer comparison study of
The authors thank CSIRO for permission to publish the research
hybrid and non-hybrid ground source heat pump systems. Appl Energy
outcomes. 2016;165:919–29.
[5] Bertani R. Geothermal power generation in the world 2010–2014 update
report. Geothermics 2016;60:31–43.
Appendix A [6] Website of Geodynamics Ltd. Australia. <http://www.geodynamics.com.au/
Our-Investments/Geothermal.aspx>.
In this section the solutions for the cases when N = 1 provided. [7] Bataille A, Genthon P, Rabinowicz M, Fritz B. Modeling the coupling between
free and forced convection in a vertical permeable slot: implications for the
From Eq. (30) we obtain heat production of an Enhanced Geothermal System. Geothermics
h pffi i 2006;35:654–82.
H1in
v s
A1 ¼ 0; B1 ¼ exp  21b 2
gðU; WÞ ;
s [8] Schweitzer R, Bilgesu HI. The role of economics on well and fracture design
h pffi i ð39Þ completions of marcellus shale wells. In: Proceedings of the SPE Eastern
H1in 2v s
B2 ¼ b2 A2 ; A2 ¼ sð1b 2Þ
exp  1b gðU; WÞ : Regional Meeting (SPE ’09), September. p. 423–8.
2 [9] Lauwerier HA. The transport of heat in an oil layer caused by the injection of
hot fluid. Appl Sci Res 1955;5:145–50. http://dx.doi.org/10.1007/BF03184614.
Therefore, the analytical solution in the Laplace space for the [10] Bödvarsson GS. On the temperature of water flowing through fractures. J
temperatures in the two rock layers and in the fluid are obtained Geophys Res 1969;74(8):1987–92. http://dx.doi.org/10.1029/
ffi JB074i008p01987.
gðU;WÞ  psffi z pffi 
p
2e s
^ 2 ¼ in e
H
H1 ð1b
2Þ e  b2 e sz
þ csZ ; [11] Cheng AHD, Ghassemi A, Detournay E. Integral equation solution of heat
r sð1b2 Þ extraction from a fracture in hot dry rock. Int J Numer Anal Meth Geomech
2e
p ffis pffi 2001;25(13):1327–38. http://dx.doi.org/10.1002/nag.182.
^1 ¼
H
H1in
e
ð1b
2Þ
gðU;WÞ  sZ
e þ csZ ; ð40Þ [12] Mohais R, Xu C, Dowd P. Fluid flow and heat transfer within a single horizontal
r s
2e
p ffis fracture in an enhanced geothermal system. J Heat Transfer 2011;133
^f ¼
H
H1in
e
ð1b Þ
gðU;WÞ
2 : (11):112603. http://dx.doi.org/10.1115/1.4004369.
s
[13] Yang S, Yeh H. An analytical solution for modeling thermal energy transfer in a
where the function g(U, W) are given below. confined aquifer system. Hydrogeol J 2008;16:1507–15. http://dx.doi.org/
10.1007/s10040-008-0327-9.
When the depth H of the bottom fracture is large enough,
[14] Gringarten AC, Sauty JP. A Theoretical study of heat extraction from aquifers
b2 0, from (40) we obtain with uniform regional flow. J Geophys Res 1975;80(35):4956–62. http://dx.

H1in pffiffi pffiffi  doi.org/10.1029/JB080i008p01120.
H2r ¼ exp 2e sgðU; WÞ þ sZ þ csZ ðZ 6 0Þ;
s
[15] Schulz R. Analytical model calculations for heat exchange in a confined
  ð41Þ aquifer. Geophys J Int 1987;61:12–20.
^ 1 ¼ Hin exp 2epffiffisgðU; WÞ  pffiffisZ þ cZ ðZ P 0Þ:
1
H [16] Wu B, Zhang X, Bunger AP, Jeffrey RG. An efficient and accurate approach for
r s s
studying the heat extraction from multiple recharge and discharge wells. In:
where the function g(U, W) is defined as follows Bunger AP, McLennan J, Jeffrey RG, editors. Effective and sustainable hydraulic
fracturing. InTech; 2013. p. 929–44.
Z þ1 [17] Gringarten AC, Witherspoon PA, Ohnishi Y. Theory of heat extraction from
df
gðU; WÞ ¼ 2
; or fractured hot dry rock. J Geophys Res 1975;80(8):1120–4. http://dx.doi.org/
U ðcoshf  cos WÞ 10.1029/JB080i008p01120.
Z q [18] Fox DB, Sutterd DB, Beckers KF, Lukawski MZ, Kocha DL, Andersone BJ, Testera
4fdf
gðq; WÞ ¼ 2
where q ¼ eU : ð42Þ JW. Sustainable heat farming: modeling extraction and recovery in discretely
0 ð1  2f cos W þ f2 Þ fractured geothermal reservoirs. Geothermics 2013;46:42–54.
[19] Wu B, Zhang X, Jeffrey RG, Bunger AP, Huddlestone-Holmes C. Perturbation
According to Gradshteyn and Ryzhik [54], the following equa- analysis for predicting the temperatures of water flowing through multiple
parallel fractures in a rock mass. Int J Rock Mech Min Sci 2015;76:162–73.
tions are valid http://dx.doi.org/10.1016/j.ijrmms.2015.03.013.
R R [20] Pruess K, Oldenburg CM. TOUGH2 user’s guide version 2. Lawrence Berkeley
1
ðcosh xþaÞ2
dx ¼ ða21
1Þ
sinh x
ðcosh xþaÞ
þ ða2a1Þ dx
ðcosh xþaÞ
; when a2 –1; Laboratory.
R h i [21] Diersch HG. FEFLOW: finite element modeling of flow, mass and heat transport
a2
1
ðcosh xþaÞ2
dx ¼ 3a sinh x þ ðcoshaxþaÞ ;
ðcosh xþaÞ2
when a ¼ 1; in porous and fractured media. Berlin Heidelberg: Springer-Verlag; 2014.
R   [22] Dempsey D, Kelkar S, Davatzes N, Hickman S, Moos D. Numerical modeling of
dx
cosh xþcos b
¼ 2
sin b
arctan tanh 2x tan 2b ; when j cos bj–1; injection, stress and permeability enhancement during shear stimulation at
R the Desert Peak Enhanced Geothermal System. Int J Rock Mech Min Sci
dx
cosh xþcos b
¼ cos b coshsinh
xcos b
x
; when j cos bj ¼ 1: 2015;78:190–206.
ð43Þ [23] Schlumberger. Eclipse, reference manual 2008.1. Technical report,
Schlumberger Informations Solutions, Business Development Central and
Set a = cos b = cos(p + W) for Eq. (43). If |a| – 1, then Eastern Europe, Hannover, Germany.
[24] Jiang F, Luo L, Chen J. A novel three-dimensional transient model for
h  i subsurface heat exchange in enhanced geothermal systems. Int Commun
gðU; WÞ ¼ sin12 W þ 2 cos W
arctan tanh 2x cotan W2 jx!þ1
sinh x
ðcosh xcos WÞ
sin W x¼U Heat Mass Transfer 2013;41:57–62.
n    [25] Llanos EM, Zarrouk S, Hogarth RA. Numerical model of the Habanero
¼ sin2 W 1  ðcosh Ucos WÞ þ 2 cot W arctan cot 2
1 sinh U W
geothermal reservoir, Australia. Geothermics 2015;53:308–19.
  [26] Hecht-Méndez J, de Paly M, Beck M, Bayer P. Optimization of energy extraction
 arctan tanh U2 cot W2 ; for vertical closed-loop geothermal systems considering groundwater flow.
Energy Convers Manage 2013;66:1–10.
[27] Biagi J, Agarwal R, Zhang Z. Simulation and optimization of enhanced
and if |a| = 1, then geothermal systems using CO2 as a working fluid. Energy 2015;86:627–37.
h i [28] Casasso A, Sethi R. Efficiency of closed loop geothermal heat pumps: a
2W x!þ1
gðU; WÞ ¼  cos3 W sinh x ðcoshcos
xcos WÞ 2  ðcosh xcos WÞ jx¼U
cos W
sensitivity analysis. Renewable Energy 2014;62:737–46.
  [29] Blöcher MG, Zimmermann G, Moeck I, Brandt W, Hassanzadegan A, Magri F.
¼ 13 1 þ coshsinh
Ucos W fcosh Ucos W  1g :
U cos W
3D numerical modeling of hydrothermal processes during the lifetime of a
deep geothermal reservoir. Geofluids 2010;10:406–21.
 B. Wu et al. / Applied Energy 183 (2016) 1664–1681 1681

[30] Hadgu T, Kalinina E, Lowry TS. Modeling of heat extraction from variably Mechanics/Geomechanics Symposium, San Francisco, CA, USA, June 26–29,
fractured porous media in Enhanced Geothermal Systems. Geothermics 2011. Paper ARMA 11–318.
2016;61:75–85. [42] Jeffrey RG, Chen ZR, Zhang X, Bunger AP, Mills KW. Measurement and analysis
[31] Deo M, Roehner R, Allis R, Moore J. Modeling of geothermal energy production of full-scale hydraulic fracture initiation and orientation. Rock Mech Rock Eng
from stratigraphic reservoirs in the Great Basin. Geothermics 2014;51:38–45. 2015;48(6):2497–512.
[32] Tenma N, Yamaguchi T, Zyvoloski G. The Hijiori Hot Dry Rock test site, Japan: [43] Bunger AP, Zhang X, Jeffrey RG. Parameters effecting the interaction among
evaluation and optimization of heat extraction from a two-layered reservoir. closely-spaced hydraulic fractures. SPE J 2012;17(1):292–306.
Geothermics 2008;37(1):19–52. [44] Strack ODL. Groundwater mechanics. Englewood Cliffs (NJ): Prentice Hall;
[33] Ghassemi A, Tarasovs S, Cheng AHD. Integral equation solution of heat 1989.
extraction-induced thermal stress in enhanced geothermal reservoirs. Int J [45] Lowell RP. Comments on ‘Theory of heat mathematical model of a hot dry rock
Numer Anal Meth Geomech 2005;29:829–44. http://dx.doi.org/ system extraction from fractured hot dry rock’ by AC Gringarten, PA
10.1002/nag.440. Witherspoon, and Yuzo Ohnishi’. J Geophys Res 1976;81:359–60. http://dx.
[34] Kumar D, Gutierrez M. Three-dimensional heat flow model for enhanced doi.org/10.1029/JB081i002p00359.
geothermal systems using boundary element method. In: Proceedings of [46] Abé H, Sekine H, Shibuya Y. Thermoelastic analysis of a cracklike reservoir in a
Thirty-Eighth Workshop on Geothermal Reservoir Engineering, Stanford hot dry rock during extraction of geothermal energy. J Energy Res Technol
University, Stanford, California, February 11–13. 1983;105:503–8. http://dx.doi.org/10.1115/1.3230963.
[35] Safari R, Ghassemi R. Three-dimensional poroelastic modeling of injection [47] Gnielinski V. New equations for heat and mass transfer in turbulent pipe and
induced permeability enhancement and microseismicity. Int J Rock Mech Min channel flow. Int Chem Eng 1976;16:359–68.
Sci 2016;84:47–58. [48] Churchill SW. A comprehensive correlating equation for lamina, assisting,
[36] McClure MW, Horne RN. An investigation of stimulation mechanisms in force and free convection. AIChE J 1977;23:10–6. http://dx.doi.org/10.1002/
Enhanced Geothermal Systems. Int J Rock Mech Min Sci 2014;72:242–60. aic.690230103.
[37] Dargush GF, Banerjee PK. Application of the boundary element method to [49] Zhao Z. On the heat transfer coefficient between rock fracture walls and
transient heat conduction. Int J Numer Meth Eng 1991;31:1231–47. flowing fluid. Comput Geotech 2014;59:105–11. http://dx.doi.org/10.1016/
[38] Geodynamics Limited. Habanero geothermal project field development plan. j.compgeo.2014.03.002.
Document Number: COM-FN-OT-PLN-01166; 2014. [50] Jiji LM. Heat conduction. 3rd ed. Verlag Berlin Heidelberg: Springer; 2009.
[39] Klee G, Bunger A, Meyer G, Rummel F, Shen B. In situ stresses in borehole [51] Press WH, Teukolsky SA, Vetterling WT, Flannery BP. Numerical Recipes in
Blanche-1/South Australia derived from breakouts, core discing and hydraulic Fortran 77: the Art of Scientific Computing. 2nd ed. Cambridge University
fracturing to 2 km depth. Rock Mech Rock Eng 2011;44:531–40. Press; 1992.
[40] Jeffrey RG, Bunger AP, Lecampion B, Zhang X, Chen ZR, van As A, et al. [52] Stehfest H. Numerical inversion of Laplace transforms. Commun ACM
Measuring hydraulic fracture growth in naturally fractured rock. In: 2009 SPE 1970;13:47–9.
Annual Technical Conference and Exhibition, 4-7 October 2009, New Orleans, [53] Cheng AHD, Sidauruk P, Abousleiman Y. Approximate inversion of the laplace
Louisiana. SPE 124919; 2009. transform. Math J 1994;4(2):76–822.
[41] Bunger AP, Jeffrey RG, Zhang X. Experimental Investigation of the Interaction [54] Gradshteyn IS, Ryzhik IM. Table of integrals, series, and
among closely spaced hydraulic fractures. In: Proceedings 45th US Rock products. Amsterdam: Academic Press; 1996.