International Journal of Heat and Mass Transfer 126 (2018) 1083–1093

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt

Heat extraction mechanism in a geothermal reservoir with rough-walled

fracture networks
Yun Chen b, Guowei Ma a,b,⇑, Huidong Wang b,c
a
School of Civil and Transportation Engineering, Hebei University of Technology, 5340 Xiping Road, Beichen District, Tianjin 300401, China
b
School of Civil, Environmental and Mining Engineering, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australia
c
College of Architecture and Civil Engineering, Beijing University of Technology, Beijing 100124, China

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

Article history: This study aims at understanding the mechanism of heat extraction from a geothermal reservoir charac-
Received 28 February 2018 terized by rough-walled fracture networks. A unified pipe-network method (UPM) which simplifies both
Received in revised form 24 April 2018 fractures and the rock matrix as pipes is developed considering the local thermal non-equilibrium (LTNE)
Accepted 20 May 2018
theory, and it is verified against an analytical solution. Three-dimensional simulations of macroscopic
Available online 29 May 2018
fluid flow and heat transfer in a fractured geothermal reservoir are conducted to take account of fracture
roughness. The channeling effect and the heterogeneous distribution of fluid temperature in a core-scale
Keywords:
model with a rough-walled fracture surface are simulated. An equivalent heat transfer coefficient (EHTC)
Equivalent heat transfer coefficient
Heat extraction
is obtained from numerical experiments with respect to the flow rate, mechanical aperture and the
Unified pipe-network method equivalent hydraulic aperture. A representative element volume is then used to investigate the flow
Geothermal reservoir and heat transfer process in a geothermal reservoir with rough-walled fracture networks by applying
Rough-walled fractures the obtained EHTC. Results demonstrate that it is essential to use the proposed EHTC since the constant
heat transfer coefficient (HTC) recommended in previous studies underestimates the final outlet fluid
temperature in cases with rough-walled fractures.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction the fluid temperature in fractures and the solid temperature along
fracture walls cannot be locally identical if the fracture spacing is
Geothermal energy is a renewable and clean energy that is more than two to three meters [3]. Therefore, a local thermal
mainly stored in the enhanced geothermal system (EGS) with com- non-equilibrium (LTNE) model [1,16–18] is introduced in the sim-
plex fracture networks embedded in low permeable rock matrix ulation of geothermal development by incorporating a heat trans-
[1,2]. It is important to evaluate the thermal production efficiency fer term controlled by the fluid-solid temperature difference,
and potential of a geothermal reservoir [3,4]. Previous experiments contact area at the interface and the heat transfer coefficient
have been performed to reveal multiphysical-interaction mecha- (HTC) [4,19–21]. The HTC determines the intensity of the heat
nisms in the process of heat extraction [5–7]. Still, numerical mod- transfer between the fluid and solid phases and thus is regarded
elling is an effective tool in estimating heat recovery in geothermal as a critical parameter in the estimation of thermal production.
reservoirs [8–11], and a local thermal equilibrium (LTE) theory is Numerous work including experimental and theoretical studies
assumed in the classical heat transfer of fluid-saturated models has been carried out to investigate the heat transfer coefficient.
[12,13]. The fluid and its surrounding matrix are considered as a Based on the convectional core-scale triaxial test platform with a
single continuum, and the heat reaches the local equilibrium cylindrical rock specimen installed, a series of flow-through exper-
instantaneously. The LTE model improves the computational effi- iments were performed by injecting the cool fluid into a man-made
ciency with an identical temperature shared by both fluid and fracture of the rock specimen which was continuously heated
matrix requiring to be calculated. Nevertheless, studies indicate [22,23]. The influence of different fracture apertures and various
that this model is not applicable if rapid heating or cooling process flow velocities on the heat transfer effect was analyzed. Ogino
exists in the formation with multiple fractures [14,15]. Especially, et al. [24] determined an average heat transfer coefficient by using
a physical model where water was injected into the two-
dimensional circular fracture plane. Large scale tests were per-
⇑ Corresponding author at: School of Civil and Transportation Engineering, Hebei
formed in Lu and Xiang [25] to investigate the effect of fluid flow
University of Technology, 5340 Xiping Road, Beichen District, Tianjin 300401, China
E-mail address: guowei.ma@uwa.edu.au (G. Ma). on the temperature distribution considering several pre-set

https://doi.org/10.1016/j.ijheatmasstransfer.2018.05.103
0017-9310/Ó 2018 Elsevier Ltd. All rights reserved.
1084 Y. Chen et al. / International Journal of Heat and Mass Transfer 126 (2018) 1083–1093

apertures in the granite. The most recent work carried out by Bai for each fracture in fracture networks with complex geometries is
et al. [26] focused on characterizing the overall HTC. Flow and heat another difficult task to be solved.
transfer tests were performed using a newly developed multi-field An equivalent parameter analysis strategy can be the effective
triaxial test system to precisely estimate thermal properties of rock solution for the aforementioned difficulties. In this study, the fluid
specimens under different confining pressure, confining tempera- flow and heat transfer in both core-scale and representative ele-
ture and fluid flow rate. Analytical models were also proposed on mentary volume (REV)-scale models are simulated using a unified
the establishment of experimental findings. A thermal boundary pipe-network method (UPM). This numerical method incorporat-
layer model was suggested by Chapman [27] to calculate the aver- ing the LTNE theory is verified against analytical solutions with dif-
age HTC for the laminar fluid flow through a flat plate maintaining ferent mesh strategies along fracture surfaces. In the core-scale
a constant temperature. However, it was demonstrated that this model with a rough-walled fracture surface, the channeling effect
model overestimated the value of the HTC according to the study and the heterogeneous distribution of fluid temperature are simu-
of Zhao [4]. In Zhao’s work, a power correlation was found to deter- lated. An equivalent heat transfer coefficient (EHTC) is quantita-
mine the HTC by combining previous experimental results [23] and tively obtained from numerical experiments with respect to the
an analytical solution for a rectangular model embedded with a flow rate, equivalent hydraulic aperture (EHA) and mechanical
single straight fissure. Explicit results of the overall HTC based on aperture. Instead of precisely describing the aperture information
a realistic half-disk model were provided in recent studies in each fracture node, equivalent parameters including the EHA
[26,28,29]. The analytical model [26] verified by the experimental and EHTC obtained are used to simulate macroscopic flow and heat
results indicated that HTC was positively correlated with the fluid transfer process in a 3D REV model embedded with rough-walled
flow rate, while the coefficient decreased with the growth of the fractures. The influences of different mechanical apertures and
fracture aperture. Different from the above models in the steady- fracture surface properties on the efficiency of heat transfer are
state condition, a transient solution was proposed to describe the investigated. The significance of considering EHTC in the modeling
local heat transfer process through a dynamic heat transfer coeffi- of geothermal development is also addressed by comparing with
cient [19]. This dynamic result was extended to simulate the heat the situation using the constant HTC suggested by previous studies
transfer in a reservoir scale. [1,3].
Most of the previous studies mentioned above were carried out
on the basis of a smooth-walled fracture model with a constant 2. Mathematical model and numerical method
fracture aperture. However, man-made or natural fractures in
EGS are typically coarse and irregular [21]. The existence of In this study, the single-phase fluid flow and heat transfer is
rough-walled fractures, on one hand, greatly influences fluid flow assumed in both rock matrix and fractures. The pressure and flow
patterns by generating the channeling effect [30,31]. Channels, rate distributions are evaluated with Darcy flow, and Cubic law is
formed by the random distribution of fracture apertures along assumed to be valid in the local area of the fracture surface. The
the fracture surface, are characterized with the tortuous property local thermal non-equilibrium (LTNE) theory is incorporated into
and have variable aperture along their length [32]. In the spatially a three-dimensional unified pipe-network method (UPM) with a
heterogeneous aperture fields, the fluid flow in these passages single fracture or multiple fractures randomly generated in the
tends to be channelized along several preferential paths. The domain. Chemical and mechanical effects are ignored in the cur-
‘‘channeling effect” is thus defined as a phenomenon of preferential rent simulation.
paths filling with an increasing portion of the fluid flow [31]. On
the other hand, the impact of fracture roughness on the heat-
2.1. Fluid flow and heat transfer
transfer mechanism between fluid and fracture walls is also
demonstrated in literatures. According to the experimental and
In the three-dimensional rock mass, both fractures and the
theoretical studies by He et al. [33], the heterogeneous distribution
matrix are considered for the fluid flow and heat transfer. The gov-
of the local HTC was mainly controlled by the degree of fracture
erning equations for pressure and temperature assuming LTNE in
roughness. Recent experiments [34] compared the convection heat
the steady state is:
transfer characteristics between distilled water flow in a man-

s 
made smooth-walled fracture and a rough-walled fracture, and k
r qf rP ¼ 0; ð1Þ
concluded that the heat transfer effect was intensified with the l
increase of the fracture roughness. Nevertheless, these qualitative
findings cannot be directly employed to the simulation of geother- qf cf r  ð! s
u  T f Þ ¼ r  ð/s kf  rT f Þ þ hint ðT s  T f Þ; ð2Þ
mal development. And few studies have been done to find the
quantitative relationship between the HTC and the fracture
roughness.
r  ðð1  /m Þks  rT s Þ þ hm
int ðT f  T s Þ ¼ 0; ð3Þ
It is also a great challenge to simulate the heat transfer in a where P is the fluid pressure; qf is the fluid density; s represents
three-dimensional model embedded with rough-walled fracture each medium in the model (s = f for the fracture, s = m for the
planes, even though numerous discrete fracture network (DFN) matrix), and / f is the fracture porosity, while /m is the matrix
based models were proposed to simulate engineering problems ! !
porosity; u is the flow velocity which can be expressed as u =
such as CO2 sequestration, heat mining and oil&gas development
(ks/l)rP based on Darcy’s law; l is the fluid viscosity; ks is the
by assuming that all fractures were smooth [35–42]. Grids with
intrinsic permeability, and kf equals to e2/12 for fractures according
extremely small sizes were assumed to be applicable to model
to the Cubic law, e is the hydraulic aperture [46]; Tf and Ts represent
the aperture distribution along fracture surfaces, aiming at pre-
the fluid temperature and the solid temperature, respectively; cf is
cisely describing the surface properties. The fluid flow channeling
the fluid heat capacity; kf and ks represent the fluid thermal conduc-
effect was discussed based on the small sized grid in literature s
[43–45]. However, these accurate simulations are only applicable tivity and the solid thermal conductivity, respectively; hint is the
f
to a core-scale model, since the high computational cost exists if solid-fluid interface heat transfer coefficient; hint is expressed as
extremely large numbers of elements are generated on fracture 2h/e for heat transfer along fractures [4], h is the fracture heat trans-
m
planes in a 3D macroscopic model to reflect the fracture surface fer coefficient; hint is the solid-fluid interface heat transfer coeffi-
roughness. On the other hand, the random distribution of apertures cient of the rock matrix. The term on the left side of Eq. (2)
 Y. Chen et al. / International Journal of Heat and Mass Transfer 126 (2018) 1083–1093 1085

represents the heat convection whose influence is notable in frac- The arithmetic mean value E of the aperture e(x, y) can be cal-
tures with high fluid flow velocity. The first term on the right side culated based on the mean a and the standard deviation r of the
of Eq. (2) and the first term on the left side of Eq. (3) are heat con- log-aperture B(x, y):
duction terms which take effects when the influence of heat con-

1 2
vection is weakened. E ¼ exp r ðln 10Þ2  10a ; ð7Þ
2
For the transient-state case, the rate of change term [47] is con-
sidered in both Eqs. (2) and (3): where E is interpreted as the mechanical aperture since the arith-
@T f metic mean aperture represents the average distance between
! s
/s qf cf þ qf cf r  ð u  T f Þ ¼ r  ð/s kf  rT f Þ þ hint ðT s  T f Þ; ð4Þ two fracture walls [52].
@t
An exponential form of the isotropic autocovariance function A
@T s (r) is employed to generate a random aperture field ([45,56]):
ð1  /m Þqs cs
m
¼ r  ðð1  /m Þks  rT s Þ þ hint ðT f  T s Þ; ð5Þ

@t 2r
AðrÞ ¼ r2 exp  ; ð8Þ
where qs is the rock density; cs is the rock heat capacity. k
where r is the separation lag, k is defined as the correlation length.
2.2. The unified pipe-network method
For a rectangular rough-walled fracture shown in Fig. 1, the
inlet and outlet boundaries are assigned with constant pressures.
The simulation tool of the unified pipe-network method (UPM)
The other boundaries are impermeable. Thus, the equivalent
was first developed by Ren et al. [48–50] in the application of
hydraulic aperture (EHA) eh can be evaluated as [44]:
single-phase and multiple-phases fluid flow in fractured porous

1=3
rock masses. The complex fracture networks are incorporated into 12lQ
the macroscopic system by random functions. A self-developed eh ¼ ; ð9Þ
WðDP=LÞ
mesh generation tool is employed to discretize fractures as trian-
gular elements with high mesh quality based on an advanced where Q is the volumetric flow rate obtained from the numerical
adaptive mesh method, followed by discretizing the rock matrix simulation; W is the width of the fracture perpendicular to the fluid
as tetrahedral elements based on a conforming mesh technique flow direction; L is the length of the fracture parallel to the fluid
[51]. In the UPM framework, the fluid flow and heat transfer infor- flow direction; DP is the pressure differential between the inlet
mation in each discrete element is assigned to both matrix and and outlet boundary.
fracture pipes according to a pipe-equivalent method. The detailed
description about the equivalent process for the fluid flow can be 3. Model verification
referred to Ren et al. [49]. In each matrix nodal control area, equiv-
alent conductance coefficients for all the matrix pipes connected to This section aims to verify the UPM based LTNE model with a
the node can be calculated according to a FEM-based shape func- smooth-walled fracture embedded in the core-scale rock mass. A
tion combining with the flow governing equation (Eq. (1)) simpli- cuboid shaped model with dimension 300 mm  50 mm  50
fied by using the Gauss’ divergence theorem. Similarly, the mm contains a rectangular fracture horizontally placed in the cen-
equivalent conductance coefficient for a fracture pipe can also be ter of the domain, as shown in Fig. 2. The rock matrix is assumed to
obtained with the same process, except that the intrinsic perme- be impermeable in this case, and only the fracture with a constant
ability is calculated based on the Cubic law. In the physical field aperture take effects in the situation of fluid flow and heat transfer
of the heat transfer, both fracture and matrix pipes are assigned between fluid and fracture walls. Still, the evolution of solid tem-
with equivalent thermal conductivities. The processes of the heat perature contributed by the heat conduction in the matrix is con-
convection and heat conduction are equivalent to the pipe net- sidered based on the Eq. (3).
works in the domain by incorporating the pipe flow information. In the process of designing boundary conditions, fluid tempera-
The equivalent parameters in the interface pipes shared by differ- tures Tin and T0 are assigned on the inlet and outlet boundaries,
ent media can be unified according to the superposition principle respectively. The constant fluid flow velocity u is used to describe
[49]. Finally, an implicit time scheme is employed to solve the flow the simple seepage field. Solid temperatures on the top and bottom
and heat transfer governing equations discretized in the UPM sides are also remained as constants, i.e., Ts = T0. The rest of bound-
framework. aries are set as non-flow and thermal isolation boundaries. The
fluid temperature is the same as the solid temperature in the initial
2.3. Generation of aperture field and estimation of equivalent
hydraulic aperture

The fluid flow and heat transfer through a single fracture with
variable apertures in a core-scale model is calculated in the UPM
framework. Previous studies have indicated that the fluid flow pat-
tern is sensitive to the distribution of fracture apertures in the
single-fracture model [52,53]. A log-normal distribution of fracture
apertures, verified based on laboratory and field studies [54,55], is
commonly employed to describe the fracture surface properties
[31,52,56]. The stochastically generated aperture values are
assigned to each node of the fracture surface divided into square
meshes. The log of the fracture aperture B(x, y) follows the normal
distribution along the x-y plane, since the fracture aperture e(x, y)
is assumed to be log-normally distributed in the current study:
Fig. 1. Boundary condition for a rough-walled fracture with the unidirectional fluid
Bðx; yÞ ¼ log10 eðx; yÞ: ð6Þ flow.
1086 Y. Chen et al. / International Journal of Heat and Mass Transfer 126 (2018) 1083–1093

!
4hK r
T f ðxÞ ¼ T 0 þ ðT in  T 0 Þ exp x ; ð10Þ
uqf cf eh ð2K r þ hHÞ

where H is the height of the model. It is demonstrated that heat

conduction in the fracture can be negligible comparing with the
thermal convection under the experimental condition, and thus this
simplification is also employed in the current numerical simulation.
Two types of mesh grids for the fracture surface are used and veri-
fied respectively, i.e., the square element and the triangular element
shown in Fig. 2(a) and (b). The square-element model aims at per-
forming the numerical experiments in the core-scale domain
embedded with the roughed-walled fracture. While the
triangular-element model is designed to simulate the process of
the heat extraction in a REV scale domain by using equivalent
parameters.
Fluid temperature in the fracture along flow direction is
extracted and compared with analytical results in different values
(a) Square elements for the fracture surface of the fracture heat transfer coefficient. Results obtained from both
square-element and triangular-element meshed models are in con-
sistent with the analytical solutions, as shown in Fig. 3, demon-
strating the reliability of the UPM in the simulation of the heat
transfer in fractures. It is also indicated that the average fluid tem-
perature in the fracture increases as the heat transfer coefficient
growths large in this case study.

4. Numerical inversion for the equivalent heat transfer

coefficient

The verified UPM is employed in this section to simulate the

channeling effect and heterogeneous distribution of temperature
along a rough-walled fracture surface in a core-scale model. Corre-
lations are obtained for the quantitative evaluation of the EHTC
based on numerical experiments combining with an analytical
solution.
(b) Triangular elements for the fracture surface
4.1. Channeling flow and heterogeneous temperature distribution
Fig. 2. Core-scale model embedded with a single smooth fracture.
As mentioned in Section 2, fracture apertures follow the two-
dimensional log-normal distribution. Three groups of rough-
condition Ts = Tf = T0 (The fluid temperature in the domain equals walled fracture surfaces are generated based on three constants
to the outlet fluid temperature at t = 0). Parameters used in this of the mechanical aperture E, namely, 20 lm, 35 lm, and 50 lm
case are listed in Table 1. for each group. Therefore, the mean a of the log-aperture B(x, y)
It should be noted that no analytical solution can be found for can be interpreted according to Eq. (7) if the value of the standard
the above three-dimensional finite system. However, an analytical deviation r is provided. In each group with the same mechanical
result for a simplified two-dimensional model (obtained by cutting aperture, rough-walled fracture surfaces are divided into five
along x-z plane at y = 25 mm in Fig. 2) with a linear fracture hori-
zontally embedded is also valid for the verification, since the 3D
model is symmetric about x-z plane at y = 25 mm. The boundary
conditions and parameters for this 2D model are the same as those
mentioned above, except that the constant outlet temperature is
assigned at x = +1. The analytical solution for the fluid tempera-
ture in the fracture can be written as [4]:

Table 1
Simulation parameters for model verification.

Parameters Symbol Unit Value

Injection velocity u mm/s 10.63
Fluid density qf kg/m3 1000
Fluid viscosity l Pas 0.001
Thermal conductivity of fluid kf J/(mKs) 0.6
Thermal conductivity of rock ks J/(mKs) 3.5
Heat capacity of fluid cf J/(kgK) 4200
Heat capacity of rock cs J/(kgK) 790
Equivalent hydraulic aperture eh lm 19.17
Inlet fluid temperature Tin K 315
Initial temperature T0 K 363 Fig. 3. Comparison of the UPM results (two types of grids) with the analytical
solution.
 Y. Chen et al. / International Journal of Heat and Mass Transfer 126 (2018) 1083–1093 1087

subgroups by different standard deviations varying from 0.3 to 0.7. between 1.0 and 2.0 [4]. In the current simulation, both A and B
The spatial correlation length is kept as a constant (k = 5 mm) for are assumed to be 1 since linear relationship between h and u is
all the subgroups. The stochastic distributions of fracture apertures obtained in the most recent experimental study [34].
are simulated based on a specific set of fracture aperture parame- The Eq. (11) together with the Cubic law is incorporated into
ters (a, r, k) [31]. And ten realizations in terms of the same set of UPM framework to calculate the heterogeneous fluid flow and heat
fracture aperture parameters are generated to investigate the transfer along the rough-walled fracture. The model employed is
dependence of realizations. the same as Fig. 2(a) using square grids, except that various aper-
In the core-scale model embedded with a single smooth frac- tures are assigned to each node rather than adopting a constant
ture, the empirical relationship between fracture heat transfer aperture along the whole fracture surface. Based on the hydraulic
coefficient and flow velocity is obtained by combining the analyt- and thermal properties listed in Table 1 combining with the
ical solution for the LTNE model with previous experimental boundary information, the pressure and temperature fields on
results [4]. This correlation is assumed to be valid in this study, the rough-walled fracture surface can be obtained for each group.
and can be expressed as: Figs. 4 and 5 are examples of simulated results containing the
heterogeneous distributions of the fracture aperture, hydraulic

B
h u pressure, flow rate and fluid temperature based on the same
¼A ; ð11Þ
h0 u0 mechanical aperture E = 20 lm.
In the case with low standard deviation (r = 0.3), the distribu-
where h is the fracture heat transfer coefficient and u is the flow tion of fracture apertures (shown in Fig. 4(a)) is more uniform than
velocity. h0 and u0 are constants introduced from experimental the case with higher standard deviation (r = 0.5) as shown in Fig. 5
results. Here, h0 is 12.64 W/(m2K) and u0 is 10.63 mm/s based on (a). The fluid channeling effect can be observed in both cases with
the first set of experiment in Zhao’s study [4,23]. Q0 which will be multiple flow passages existing in the region featuring with large
used in Section 4.2 is also obtained from this experiment, and set aperture value. Relatively higher fluid flow rate is obtained when
as 10.39 mm3/s. Furthermore, A is around 1.0, and B is suggested the standard deviation is low by comparing Fig. 4(b) with Fig. 5

(a) Distribution of aperture (b) Distribution of flow rate

(c) Distribution of pressure (d) Distribution of temperature

Fig. 4. Representative results for the fluid flow and heat transfer within a rough-walled fracture (E = 20 lm, r = 0.3).

(a) Distribution of aperture (b) Distribution of flow rate

(c) Distribution of pressure (d) Distribution of temperature

Fig. 5. Representative results for the fluid flow and heat transfer within a rough-walled fracture (E = 20 lm, r = 0.5).
1088 Y. Chen et al. / International Journal of Heat and Mass Transfer 126 (2018) 1083–1093

(b). It is difficult for the fluid to pass along the fracture surface with
intensified heterogeneity of the aperture distribution. Flow pas-
sages are narrowed resulting in a small amount of fluid passing
through the whole rough surface, as shown in Fig. 5(b). The hetero-
geneous distributions of pressure and temperature for both cases
are represented in Figs. 4(c), (d) and 5(c), (d). Different from the
smooth-surface situation where isolines for pressure or tempera-
ture values are parallel with the inlet boundary, the isolines are
twisted because of the unevenly distributed apertures. The distri-
bution of fluid temperature, particularly, is more heterogeneous
in the case with the high value of r, as shown in Fig. 5(d). On the
other hand, the lower fluid temperature extends further from the
inlet boundary (shown in Fig. 4(d)) if the standard deviation is
low. This phenomenon is generated by the large flow rate combin-
ing with the intensity of the heat transfer between the fluid and
fracture walls.

4.2. Equivalent heat transfer coefficient

On the establishment of the above numerical experiments, the

fluid flow rate at the outlet boundary, the EHA as well as the fluid
temperature at the location x = 0.1 m are calculated for each case.
(a) Discretizing fractures as triangular elements
These parameters are employed to solve the EHTC combining with
the analytical solution (Eq. (10)) mentioned in Section 3. Eq. (10)
can also be written as:

2kr uqf cf eh M
heq ¼  ; ð12Þ
4xkr þ uqf cf MH
 
T f ðxÞT f 0
where M ¼ ln T in T f 0
.
The average values of the ratio of the EHA to the mechanical
aperture eh/E and the dimensionless ratio of the heat transfer coef-
ficient to the fluid flow rate (heq/h0)/(Q/Q0) are calculated for each
subgroup corresponding with specific standard deviation and
mechanical aperture. It should be noted that the simulated results
have minor dependence on realizations in each subgroup. The
maximum percentage of dispersion is about 10% for the value of
(heq/h0)/(Q/Q0) in the case with high standard deviation (r = 0.7).
On the other hand, the value of eh/E remains at an identical interval
if the standard deviation is the same, indicating that this term is
only r dependent. The relationship between these two average
ratios are plotted in Fig. 6 for the correlation analysis, and the
power function form is found suitable for the fitting process. Three
correlations are obtained for each case with specific mechanical
aperture E:
(b) Discretizing matrix as tetrahedral elements
Fig. 7. Distribution of 100 fractures in the rock mass.

Table 2
Parameters for stochastically generating fracture networks.

Normal distribution
Item SubItem Mean Standard deviation
Fracture size Length of major axis 1.5 (m) 1
Length of minor axis 1.5 (m) 1
Fisher distribution (Fisher coefficient is 15)
Item Mean Minimum Maximum
Dip direction 30° (Group 1) 0° 60°
210° (Group 2) 180° 240°
Dip 45° 0° 90°

Fig. 6. Correlation diagram between eh/E and (heq/h0)/(Q/Q0).

 Y. Chen et al. / International Journal of Heat and Mass Transfer 126 (2018) 1083–1093 1089

Table 3 When E ¼ 20 lm; ðheq =h0 Þ=ðQ =Q 0 Þ ¼ 0:81ðeh =EÞ0:51 : ð13Þ

Parameters for macroscopic flow and heat transfer in a REV-scale model.

Parameters Symbol Unit Value

When E ¼ 35 lm; ðheq =h0 Þ=ðQ =Q 0 Þ ¼ 0:51ðeh =EÞ0:54 : ð14Þ
Pressure difference DP MPa 1
Intrinsic permeability of matrix km m2 1  1018
Porosity of matrix /m 0.01
When E ¼ 50 lm; ðheq =h0 Þ=ðQ =Q 0 Þ ¼ 0:36ðeh =EÞ0:59 : ð15Þ
Porosity of fracture /f 1
3
The solid-fluid interface hint W/(m K) 10 It should be mentioned that the degree of the fracture surface
heat transfer coefficient roughness can be qualitatively represented by the ratio of the
Initial temperature T0 K 473
EHA and the mechanical aperture. The decrease of eh/E indicates
Inlet fluid temperature Tin K 315
the intensified degree of the fracture roughness [52,57]. From

Time = 10 days Time = 30 days Time = 50 days

x-z

y-z

(a) Fluid temperature, K (E = 35 μm, smooth-walled fracture networks)

x-z

y-z

(b) Fluid temperature, K (E = 35 μm, rough-walled fracture networks)

Fig. 8. Distribution of fluid temperature at different simulation time (model with smooth-walled fractures vs. model with rough-walled fractures).
1090 Y. Chen et al. / International Journal of Heat and Mass Transfer 126 (2018) 1083–1093

Eqs. (13) to (15), it can be concluded that the equivalent heat trans- Three sets of mechanical apertures, 20 lm, 35 lm and 50 lm,
fer coefficient increases with the growth of fluid flow rate. Further- are used in the modeling, respectively. Fig. 8 represents the simu-
more, the intensified surface roughness also increases the value of lated results for the situation E = 35 lm in different simulation
heq when the flow rate is kept as a constant. These results are in time. Slices of x-z plane at y = 25 m and y-z plane at x = 25 m are
consistent with experimental discoveries in Li et al. [34]. selected to indicate the evolution of fluid temperature with time.
Two models considering rough-walled and smooth-walled frac-
tures respectively are simulated and compared. It is evident that
5. Simulation of geothermal development in a REV-scale model
the fluid temperature in the domain decreases with simulation
with fracture networks
time whether fractures are smooth or not. The shape of fractures
is prominent in the temperature distribution diagram when frac-
Fracture networks are considered in this section to simulate the
tures are smooth, as shown in Fig. 8(a). This is because the rela-
process of heat mining in geothermal reservoir. As shown in Fig. 7,
tively high flow rate in the smooth fracture facilitates the
a cubic model of side length 5 m is designed based on the size of
migration of the injected water with lower temperature. The rapid
representative elementary volume (REV) in a typical field case
reduction of the temperature in the whole domain, on the other
([30,58]). Different from the core-scale model where the fracture
hand, is occurred due to the high value of the EHTC contributed
is discretized as fine structured square elements to perfectly
by the high flow rate. The temperature field in 50 days indicates
describe the aperture information along the fracture surface, an
that thermal energy is almost completely extracted when fractures
advanced adaptive mesh method [51] is used to discretize the
are considered as smooth. However, the fluid temperature is more
large-scale complex system with fracture networks. Equivalent
uniformly distributed in the domain if the fracture roughness is
information obtained from numerical experiments in Section 4 is
considered, as shown in Fig. 8(b). The flow rate is comparatively
employed in this case to consider the influence of the fracture
much slower due to the small value of the EHA. The newly injected
roughness on the macroscopic flow and heat transfer.
fluid retained in fractures has enough time to be heated by the sur-

5.1. The distribution of fractures and fracture equivalent hydraulic

aperture

All the fractures randomly generated in the domain are ellipti-

cal. The geometrical properties of fractures are controlled by
parameters including the length of major and minor axes, the dip
and dip direction. 100 fractures, divided into two groups with
equal number, are uniformly embedded in the microscopic model.
The distribution parameters for generating geometric information
of fractures in the simulation are listed in Table 2.
Each generated fracture is discretized as triangular elements as
shown in Fig. 7(a), while the matrix is discretized as tetrahedral
elements based on the conforming mesh method (shown in
Fig. 7(b)). The total number of elements used in this simulation
is around 1,100,000.
As mentioned in Section 4, the ratio of the EHA to the mechan-
ical aperture eh/E can be used to indicate the rough-walled proper-
ties along fracture surfaces. The fracture surface roughness varies
in fractures in the complex fracture network system, which is rep-
resented by a normal distribution of E/eh employed in the current
simulation. Distribution parameters including the mean value Fig. 9. Temperature draw-down curves with different mechanical apertures and
and the standard deviation are 7.5 and 1, respectively. Therefore, fracture surface properties.

the EHA can be derived based on E/eh for each fracture. It should
be noted that the value of EHA is valid only in the condition E/eh
> 1.
In each fracture node, the equivalent heat transfer coefficient
heq is no longer a constant. The local heq can be calculated based
on Eqs. (13)-(15) combining with the local hydraulic parameters
related with eh.

5.2. Macroscopic flow and heat transfer in a REV-scale model

In a REV-scale model as shown in Fig. 7, a constant fluid temper-

ature is assigned at the bottom boundary (inlet), while the temper-
ature gradient is kept as zero on the top (outlet). The rest
boundaries are impermeable and thermal-isolated. The pressure
difference is kept as a constant between the inlet and outlet side.
The fluid and solid temperatures are T0 in the initial situation.
Parameters used in the simulation are listed in Table 3. The fluid
flow properties and other thermal information including conduc-
tivities and heat capacities for both fluid and solid phases can be
referred to Table 1. The total simulation time is 100 days in this Fig. 10. Temperature draw-down curves with different heat transfer coefficients (E
case. = 50 lm, rough-walled fractures).
 Y. Chen et al. / International Journal of Heat and Mass Transfer 126 (2018) 1083–1093 1091

rounding matrix maintaining with much higher temperature. The flow rate dominates the heat transfer effect in the geothermal
overall temperature is still within a high level even after 50 days. development. The higher flow rate increases the value of EHTC
The average fluid temperature at the outlet boundary is calcu- and intensifies the heat transfer between the fluid in fractures
lated in each time step [1], which contributes to the final outlet and fracture walls. The thermal energy reserved in the rock mass
temperature draw-down curves for cases with the different is extracted in relatively short period by the continuous rapid fluid
mechanical aperture and the fracture surface property, as shown flow.
in Fig. 9. For each group with the same mechanical aperture, it is The sensitivity analysis in respect of the HTC is carried out for
indicated that the outlet temperature in the situation with this multi-fractured model. For the situation with rough-walled
smooth-walled fractures decreases sharply and becomes much fractures embedded, the mechanical aperture is kept as 50 lm,
lower by comparing with the case where fracture roughness is con- and different values of the ETC are used (h = 10, 100, 500, 5000
sidered. The intensified flow impedance resulted from the fracture W/(m2K)) to simulate the heat mining process. The average fluid
roughness prevents the rapid heat production rate, maintaining temperature at the outlet boundary is calculated and the outlet
the outlet fluid temperature at a high level. In the comparison temperature draw-down curves are plotted in Fig. 10. It is
among three groups with the same fracture surface properties, indicated that the decrease rate of the outlet temperature acceler-
thermal breakthrough takes place at much earlier time when the ates with the growth of HTC, but it does not change much when h
mechanical aperture is 50 lm. It is demonstrated that the fluid is larger than 100 W/(m2K) since the sufficient heat transfer effect

Time = 10 days Time = 30 days Time = 50 days

(a) Fluid temperature, K

(use EHTC, E = 50 μm for the rough-walled fracture system)

(b) Fluid temperature, K

(use the constant HTC 100 W/(m2·K), E = 50 μm for the rough-walled fracture system)
Fig. 11. Distribution of fluid temperature at different simulation time (model with heq influenced by hydraulic parameters vs. model with a constant h).
1092 Y. Chen et al. / International Journal of Heat and Mass Transfer 126 (2018) 1083–1093

contributes to the local thermal equilibrium between the fluid performed based on the equivalent parameters obtained from
temperature in fractures and the temperature in fracture walls. core-scale simulations. This method simplifies the process of rep-
The heat transfer process will not continue even if h becomes resenting the fracture roughness in a larger-scale domain and
extremely large. Therefore, the outlet temperature approaches improves the computational efficiency. It is concluded that
the same value after 100 days. Results obtained based on different rough-walled fractures with lower permeability reduce the flow
value of HTC are also compared with the situation incorporating rate and slow down the cooling process of the rock mass in the
the EHTC which is locally influenced by the flow rate and the frac- condition with the same mechanical aperture, and thus the ther-
ture apertures. It is concluded that the decrease rate of the outlet mal breakthrough time is relatively longer. In the situation with
temperature in the model with parameter-dependent heq is even the same fracture roughness information, the increase of the
slower than the case where h remains 10 W/(m2K) after 30 days. mechanical aperture facilitates the decrease rate of the outlet tem-
The constant value of HTC is much larger in previous studies, perature. Furthermore, it is also demonstrated that the constant
e.g., ranging from 300 to 900 W/(m2K) in Shaik et al. [3] and equal- value of the heat transfer coefficient recommended in previous
ing to 3000 W/(m2K) in Sun et al. [1]. Therefore, the outlet temper- studies will underestimate the final outlet fluid temperature in
ature obtained by employing the HTC in previous work is about 25 the simulation of geothermal development with rough-walled
K lower than the situation that the EHTC is used in the simulation fracture networks. Thus, the EHTC proposed in the current study
after 100 days, leading to the underestimation of the final heat is essential and should be employed in the engineering analysis.
production. Several limitations should be addressed in this study. The
Fig. 11 represents the distribution of fluid temperature in differ- geothermal development is a multiphysical coupling process
ent simulation time for both cases where the heq is influenced by where chemical and mechanical effects should be taken into con-
hydraulic parameters and the h remains 100 W/(m2K). The aver- sideration in the future study. Non-Darcy fluid flow in fractures
age value of the parameter-dependent heq is comparatively low is also demonstrated in previous studies and should be discussed
as indicated in Fig. 11(a). The weak effect of heat transfer keeps and incorporated to develop a more powerful simulation tool.
the temperature in the matrix at a high level, which makes the
fractures containing fluid with low temperature prominent in the
temperature field. However, the intensified heat transfer effect Conflict of interest
leads to the phenomenon of the local thermal equilibrium in the
case that h equals to 100 W/(m2K), as shown in Fig. 11(b). The The authors declared that there is no conflict of interest.
fluid temperature in fractures is almost the same as that in sur-
rounding matrix at each time step. This sufficient heat transfer pro-
cess contributes to a much lower outlet temperature as Acknowledgements
represented in Fig. 11(b) when the simulation time is 50 days.
This research is supported by the Research Training Program
(RTP) Scholarship and the University Postgraduate Award for Inter-
6. Conclusion national Students (UPAIS) from the Australian Government, as well
as the Top-Up Scholarship from the University of Western
A unified pipe-network method is proposed to simulate the Australia.
fluid flow and heat transfer in the local thermal non-equilibrium
model by considering rough-walled fractures embedded in the References
rock mass. Both fractures and the matrix are discretized as pipes,
which facilitates the coupling process of fracture pipes and matrix [1] Z.X. Sun, X. Zhang, Y. Xu, J. Yao, H.X. Wang, S. Lv, Z.L. Sun, Y. Huang, M.Y. Cai, X.
pipes by a simple superposition method. Two sets of mesh gener- Huang, Numerical simulation of the heat extraction in EGS with thermal-
hydraulic-mechanical coupling method based on discrete fractures model,
ating methods are used in a single fracture system and verified Energy 120 (2017) 20–33.
against analytical solutions. The square element system can be [2] L.F. Fan, Z.J. Wu, Z. Wan, J.W. Gao, Experimental investigation of thermal
used to simulate the channeling effect and the heterogeneous tem- effects on dynamic behavior of granite, Appl. Therm. Eng. 125 (2017) 94–103.
[3] A.R. Shaik, S.S. Rahman, N.H. Tran, T. Tran, Numerical simulation of fluid-rock
perature field by precisely describing the aperture distribution
coupling heat transfer in naturally fractured geothermal system, Appl. Therm.
along the fracture surface in a core-scale model. On the other hand, Eng. 31 (10) (2011) 1600–1606.
the triangular element system combining with the conforming [4] Z.H. Zhao, On the heat transfer coefficient between rock fracture walls and
mesh technique can be used in a REV-scale simulation considering flowing fluid, Comput. Geotech. 59 (2014) 105–111.
[5] R.B. Schmidt, R. Seithel, K. Bucher, I. Stober, Fluid-rock interaction in deep fault
randomly generated fracture networks. systems and the influence on permeability in typical rocks of the Upper Rhine
In the core-scale model, numerical experiments about the flow Graben, southwest Germany, in: Proceedings World Geothermal Congress.
and heat transfer through a single rough-walled fracture are per- Melbourne, Australia, 2015.
[6] G. Barla, Comprehensive study including testing, monitoring and thermo-
formed in the UPM framework. The simulated results indicated hydro modelling for design and implementation of a geothermal system in
that the equivalent hydraulic aperture (EHA) decreases as the stan- Torino (Italy), Geomech. Geophys. Geo. Energy. Geo. Res. 3 (2) (2017) 175–188.
dard deviation of the log-normal distribution model for apertures [7] W.G.P. Kumari, P.G. Ranjith, M.S.A. Perera, B.K. Chen, Experimental
investigation of quenching effect on mechanical, microstructural and flow
becomes large, which intensifies the degree of the fracture rough- characteristics of reservoir rocks: thermal stimulation method for geothermal
ness. The flow rate is diminished in the process of fluid channeling energy extraction, J. Petrol. Sci. Eng. 162 (2018) 419–433.
and the distribution of fluid temperature becomes more heteroge- [8] W.J. Cao, W.B. Huang, F.M. Jiang, Numerical study on variable thermophysical
properties of heat transfer fluid affecting EGS heat extraction, Int. J. Heat Mass
neous. Combining the numerical results with an analytical solu- Transfer 92 (2016) 1205–1217.
tion, the equivalent heat transfer coefficient can be obtained with [9] W.B. Huang, W.J. Cao, F.M. Jiang, Heat extraction performance of EGS with
respect to the flow rate, mechanical aperture and the EHA. It is heterogeneous reservoir: a numerical evaluation, Int. J. Heat Mass Transfer 108
(2017) 645–657.
demonstrated that either increase of the flow rate or the fracture
[10] M. White, P.C. Fu, M. McClure, G. Danko, D. Elsworth, E. Sonnenthal, S. Kelkar,
roughness would promote the value of EHTC. This conclusion is R. Podgorney, A suite of benchmark and challenge problems for enhanced
in consistent with the experimental findings of Li et al. [34]. geothermal systems, Geomech. Geophys. Geo. Energy. Geo. Res. 4 (1) (2018)
The macroscopic flow and heat transfer process is also simu- 79–117.
[11] Y. Chen, G.W. Ma, H.D. Wang, T. Li, Evaluation of geothermal development in
lated in the REV-scale model embedded with rough-walled frac- fractured hot dry rock based on three dimensional unified pipe-network
ture networks. An equivalent parameter analysis strategy is method, Appl. Therm. Eng. 136 (2018) 219–228.
 Y. Chen et al. / International Journal of Heat and Mass Transfer 126 (2018) 1083–1093 1093

[12] S.N. Pandey, A. Chaudhuri, H. Rajaram, S. Kelkar, Fracture transmissivity [36] Y. Xia, Y. Jin, J. Oswald, M. Chen, K.P. Chen, Extended finite element modeling
evolution due to silica dissolution/precipitation during geothermal heat of production from a reservoir embedded with an arbitrary fracture network,
extraction, Geothermics 57 (2015) 111–126. Int. J. Numer. Methods Fluids. 86 (2018) 329–345.
[13] S.N. Pandey, A. Chaudhuri, The effect of heterogeneity on heat extraction and [37] C. Xu, P.A. Dowd, Z.F. Tian, A simplified coupled hydro-thermal model for
transmissivity evolution in a carbonate reservoir: a thermo-hydro-chemical enhanced geothermal systems, Appl. Energy 140 (2015) 135–145.
study, Geothermics 69 (2017) 45–54. [38] Z.J. Wu, L.F. Fan, Q.S. Liu, G.W. Ma, Micro-mechanical modeling of the macro-
[14] J.J. Gao, J.G. Deng, K. Lan, Y.T. Feng, W.D. Zhang, H.D. Wang, Porothermoelastic mechanical response and fracture behavior of rock using the numerical
effect on wellbore stability in transversely isotropic medium subjected to local manifold method, Eng. Geol. 225 (2017) 49–60.
thermal non-equilibrium, Int. J. Rock Mech. Min. Sci. 96 (2017) 66–84. [39] G.W. Ma, H.D. Wang, L.F. Fan, B. Wang, Simulation of two-phase flow in
[15] L.W. He, Z.H. Jin, Effects of local thermal non-equilibrium on the pore pressure horizontal fracture networks with numerical manifold method, Adv. Water
and thermal stresses around a spherical cavity in a porous medium, Int. J. Eng. Res. 108 (2017) 293–309.
Sci. 49 (3) (2011) 240–252. [40] Y. Chen, G.W. Ma, T. Li, Y. Wang, F. Ren, Simulation of wormhole propagation
[16] R. Gelet, B. Loret, N. Khalili, A thermos-hydro-mechanical coupled model in in fractured carbonate rocks with unified pipe-network method, Comput.
local thermal non-equilibrium for fractured HDR reservoir with double Geotech. 98 (2018) 58–68.
porosity, J. Geophys. Res.: Solid Earth 117 (B7) (2012) 1–23. [41] G.W. Ma, Y. Chen, Y. Jin, H.D. Wang, Modelling temperature-influenced
[17] F.M. Jiang, L. Luo, J.L. Chen, A novel three-dimensional transient model for acidizing process in fractured carbonate rocks, Int. J. Rock. Mech. Min. Sci. 105
subsurface heat exchange in enhanced geothermal systems, Int. Commun. (2018) 73–84.
Heat Mass Transfer 41 (2013) 57–62. [42] F. Ren, G.W. Ma, Y. Wang, L.F. Fan, H.H. Zhu, Two-phase flow pipe network
[18] X.L. Ouyang, R.N. Xu, P.X. Jiang, Three-equation local thermal non-equilibrium method for simulation of CO2 sequestration in fractured saline aquifers, Int. J.
model for transient heat transfer in porous media: the internal thermal Rock Mech. Min. Sci. 98 (2017) 39–53.
conduction effect in the solid phase, Int. J. Heat Mass Transfer 115 (2017) [43] T. Ishibashi, N. Watanabe, N. Hirano, A. Okamoto, N. Tsuchiya, Beyond-
1113–1124. laboratory-scale prediction for channeling flows through subsurface rock
[19] T. Heinze, S. Hamidi, B. Galvan, A dynamic heat transfer coefficient between fractures with heterogeneous aperture distributions revealed by laboratory
fractured rock and flowing fluid, Geothermics 65 (2017) 10–16. evaluation, J. Geophys. Res.: Solid Earth 120 (1) (2015) 106–124.
[20] W.B. Huang, W.J. Cao, J. Guo, F.M. Jiang, An analytical method to determine the [44] N. Watanabe, N. Hirano, N. Tsuchiya, Determination of aperture structure and
fluid-rock heat transfer rate in two-equation thermal model for EGS heat fluid flow in a rock fracture by high-resolution numerical modeling on the
reservoir, Int. J. Heat Mass Transfer 113 (2017) 1281–1290. basis of a flow-through experiment under confining pressure, Water Resources
[21] Y.Q. Ma, Y.J. Zhang, Z.W. Yu, Y.B. Huang, C. Zhang, Heat transfer by water Res. 44 (6) (2008) 1–11.
flowing through rough fractures and distribution of local heat transfer [45] W. Zhang, B.B. Dai, Z. Liu, C.Y. Zhou, A pore-scale numerical model for non-
coefficient along the flow direction, Int. J. Heat Mass Transfer 119 (2018) Darcy fluid flow through rough-walled fractures, Comput. Geotech. 87 (2017)
139–147. 139–148.
[22] J. Zhao, Experimental Studies of the Hydro-Thermo-Mechanical Behaviour of [46] R.W. Zimmerman, G.S. Bodvarsson, Hydraulic conductivity of rock fractures,
Joints in Granite PhD thesis, Imperial College, London, 1987. Transp. Porous Media 23 (1) (1996) 1–30.
[23] J. Zhao, C.P. Tso, Heat transfer by water flow in rock fractures and the [47] H.K. Versteeg, W. Malalasekera, An Introduction to Computational Fluid
application to hot dry rock geothermal systems, Int. J. Rock Mech. Min. Sci. Dynamics: The Finite Volume Method, Pearson Education, 2007.
Geomech. Abstr. 30 (6) (1993) 633–641. [48] F. Ren, G.W. Ma, Y. Wang, L.F. Fan, Pipe network model for unconfined seepage
[24] F. Ogino, M. Yamamura, T. Fukuda, Heat transfer from hot dry rock to water analysis in fractured rock masses, Int. J. Rock Mech. Min. Sci. 88 (2016) 183–
flowing through a circular fracture, Geothermics 28 (1) (1999) 21–44. 196.
[25] W. Lu, Y.Y. Xiang, Experiments and sensitivity analyses for heat transfer in a [49] F. Ren, G.W. Ma, Y. Wang, T. Li, H.H. Zhu, Unified pipe network method for
meter-scale regularly fractured granite model with water flow, J. Zhejiang simulation of water flow in fractured porous rock, J. Hydrol. 547 (2017) 80–96.
Univ. Sci. A 13 (12) (2012) 958–968. [50] F. Ren, G.W. Ma, L.F. Fan, Y. Wang, H.H. Zhu, Equivalent discrete fracture
[26] B. Bai, Y.Y. He, X.C. Li, J. Li, X.X. Huang, J.L. Zhu, Experimental and analytical networks for modelling fluid flow in highly fractured rock mass, Eng. Geol. 229
study of the overall heat transfer coefficient of water flowing through a single (2017) 21–30.
fracture in a granite core, Appl. Therm. Eng. 116 (2017) 79–90. [51] Y. Wang, G.W. Ma, F. Ren, T. Li, A constrained Delaunay discretization method
[27] A.J. Chapman, Heat Transfer, fourth ed., Macmillan Publishing Company, New for adaptively meshing highly discontinuous geological media, Comput.
York, 1984. Geosci. 109 (2017) 134–148.
[28] G.W. Zhang, J.L. Zhu, J. Li, Q.L. Wang, The analytical solution of the water-rock [52] C.E. Renshaw, On the relationship between mechanical and hydraulic
heat transfer coefficient and sensitivity analyses of parameters, in: apertures in rough-walled fractures, J. Geophys. Res.: Solid Earth 100 (B12)
Proceedings World Geothermal Congress. Melbourne, Australia, 2015. (1995) 24629–24636.
[29] X.X. Huang, J.L. Zhu, J. Li, B. Bai, G.W. Zhang, Fluid friction and heat transfer [53] C.F. Tsang, I. Neretnieks, Flow channeling in heterogeneous fractured rocks,
through a single rough fracture in granitic rock under confining pressure, Int. Rev. Geophys. 36 (2) (1998) 275–298.
Commun. Heat Mass Transfer 75 (2016) 78–85. [54] P.J. Bourke, D.P. Hodgkinson, E.M. Durrance, M.J. Heath, Fracture Hydrology
[30] S. Luo, Z.H. Zhao, H. Peng, H. Pu, The role of fracture surface roughness in Relevant to Radionuclide Transport. AERE Rep. 11414, Atomic Energy Res.
macroscopic fluid flow and heat transfer in fractured rocks, Int. J. Rock Mech. Estab., Harwell, United Kingdom, 1985.
Min. Sci. 87 (2016) 29–38. [55] R.A. Johns, J.S. Steude, L.M. Castanier, P.V. Roberts, Nondestructive
[31] B. Guo, P.C. Fu, Y. Hao, C.A. Peters, C.R. Carrigan, Thermal drawdown-induced measurements of fracture aperture in crystalline rock cores using X ray
flow channeling in a single fracture in EGS, Geothermics 61 (2016) 46–62. computed tomography, J. Geophys. Res.: Solid Earth 98 (B2) (1993) 1889–
[32] Y.W. Tsang, C.F. Tsang, Channel model of flow through fractured media, Water 1900.
Resource Res. 23 (3) (1987) 467–479. [56] L. Moreno, Y.W. Tsang, C.F. Tsang, F.V. Hale, I. Neretnieks, Flow and tracer
[33] Y.Y. He, B. Bai, S.B. Hu, X.C. Li, Effects of surface roughness on the heat transfer transport in a single fracture: a stochastic model and its relation to some field
characteristics of water flow through a single granite fracture, Comput. observations, Water Resources Res. 24 (12) (1988) 2033–2048.
Geotech. 80 (2016) 312–321. [57] N. Barton, S. Bandis, K. Bakhtar, Strength, deformation and conductivity
[34] Z.W. Li, X.T. Feng, Y.J. Zhang, C. Zhang, T.F. Xu, Y.S. Wang, Experimental coupling of rock joints, Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 22 (3)
research on the convection heat transfer characteristics of distilled water in (1985) 121–140.
manmade smooth and rough rock fractures, Energy 133 (2017) 206–218. [58] K.B. Min, L. Jing, O. Stephansson, Fracture system characterization and
[35] Y. Xia, Y. Jin, M. Chen, Comprehensive methodology for detecting fracture evaluation of the equivalent permeability tensor of fractured rock masses
aperture in naturally fractured formations using mud loss data, J. Petrol. Sci. using a stochastic REV approach, Hydrogeol. J. 12 (5) (2004) 497–510.
Eng. 135 (2015) 515–530.