Available online at www.sciencedirect.

com
ScienceDirect

Comput. Methods Appl. Mech. Engrg. 356 (2019) 465–489

www.elsevier.com/locate/cma

A three-dimensional thermo-hydro-mechanical coupled model for

enhanced geothermal systems (EGS) embedded with discrete fracture
networks
Yang Wanga , Tuo Lia , Yun Chena , Guowei Maa,b ,∗
a School of Civil, Environmental and Mining Engineering, The University of Western Australia, Perth, WA 6009, Australia
b School of Civil and Transportation Engineering, Hebei University of Technology, Tianjin 300401, China

Received 12 April 2019; received in revised form 24 June 2019; accepted 25 June 2019
Available online 30 July 2019

Abstract
Fractures play an important role in the geothermal energy recovery from enhanced geothermal systems (EGS) involving
interacted multi-physical fields. A strongly coupled thermo-hydro-mechanical (THM) model is proposed to simulate the
process of the long-term geothermal production in three-dimensional geothermal reservoirs containing arbitrary discrete fracture
networks (DFNs). By introducing a strong discontinuity concept into the DFN model, the aperture variation of each fracture
induced by the fluid pressure, external stresses and thermal expansion in the period of production can be captured. Non-
isothermal fluid flow in DFN and the local thermal non-equilibrium (LTNE) between fluid and rock matrix are formulated and
coupled with the fracture deformation model. Verification is carried out against an analytical solution, followed by a sensitivity
and convergence analysis concerning time step and mesh size. This approach is then applied to Habanero EGS project in
Australia to evaluate the geothermal productivity and efficiency for a period of 20 years with different injection and production
pressures. The results demonstrate that the proposed model is robust and effective to simulate the THM coupled process in
3D fractured reservoirs.
⃝c 2019 Elsevier B.V. All rights reserved.

Keywords: Enhanced geothermal systems (EGS); Discrete fracture networks (DFNs); Strongly coupled thermo-hydro-mechanical (THM) model;
Non-isothermal fluid flow; Local thermal non-equilibrium (LTNE); Habanero EGS project

1. Introduction
Geothermal energy is an attractive renewable and clean energy that are largely stored in Earth’s crust. At the end
of 2014, the world’s direct utilisation of geothermal energy has exceeded 70 GW already, which was a significant
increase by 46.2% from 2010 [1,2]. Hydro-geothermal reservoirs only began generating power commercially until
2013, in which the first private commercial EGS plant in Habanero (Australia) was launched [3]. The concepts of
hot dry rock (HDR) and enhanced geothermal system (EGS) were primarily proposed and tested at Fenton Hill
by the Los Alamos National Laboratory in the early 1970s for exploring the feasibility of utilising the abundant
∗ Corresponding author at: School of Civil, Environmental and Mining Engineering, The University of Western Australia, Perth, WA 6009,
Australia.
E-mail address: [email protected] (G. Ma).

https://doi.org/10.1016/j.cma.2019.06.037
0045-7825/⃝ c 2019 Elsevier B.V. All rights reserved.
466 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

heat energy from deep thermal reservoirs [4,5]. Hot dry rocks are tight formations that are deeply buried 3 to 10
kilometres below the earth’s surface and contain sufficient heat content. Due to the low intrinsic permeability and
porosity of HDR, it is intractable to circulate water through the high-temperature zone for recovering heat energy.
EGS is the geothermal reservoirs of HDR that are enhanced by hydraulic stimulation technologies for improving
the connectivity of fractures and the permeability of reservoirs [6–9]. Effective use of EGS can expand geothermal
resources and improve geothermal productivity and production efficiency.
The geothermal development in EGS is a complicated process involving the coupled multi-physical fields,
including the stress-field, thermal field and flow field [10–13]. Numerical simulation is an essential and efficient tool
to approach the sophisticated coupled interactions in this process. Many researchers applied continuous models to
the numerical study of the geothermal extraction process [14]. Jiang et al. [15,16] used a 3D equivalent porous model
to study the long-term heat exchange process in EGS, and they found that the triplet well layout can dramatically lift
the heat productivity through restraining the preferential flow in the reservoir. Based on this method, Cao et al. [17]
employed empirical porosity-effective stress and permeability–porosity relationships for modelling the thermos-
hydro-mechanical (THM) interaction in a porous medium. Zeng et al. [18] found that the two horizontal well
system can significantly enhance the water production rate and possibly improve the production efficiency through
implementing a series of numerical experiments for Desert Peak EGS project. Wu et al. [19] also proposed a 2D
semi-analytical model for a planar geothermal reservoir with a constant hydraulic aperture to optimise the well
placement in a dipole well system. It is found that the product temperature is independent of the well separation
if the separation is larger than some value when other parameters are fixed. As the single-porosity continuous
model does not explicitly take fractures into account, the mechanical reactions on fractures are difficult to be
incorporated into the numerical simulations. In contrast, the dual porosity method can distinguish the different
hydraulic properties of the fracture and the rock matrix and then treat them as different porous media [20,21].
Gelet et al. [22] extended the dual-porosity method into the coupled THM simulation in EGS. Besides, Taron and
Elsworth [23] developed a new simulator under the basis of the dual-porosity model to examine the permeability
variation in a prototypical geothermal reservoir responding to thermo-hydro-mechanical–chemical (THMC) fields.
Although the continuum-based methods can simplify the computation and obtain high efficiency, they often ignore
the effects of the configuration of complex fractures and the fracture connectivity on the hydraulic permeability and
thus the performance of heat extractions in EGS.
Fractures are an important factor to influence the hydraulic, mechanical and thermal characteristics of rocks
[24–32]. Being the preferential flow conduits and heat exchange zones in EGS, it is a necessary and essential
component to evaluate the performance of geothermal extraction. A realistic representation of fracture networks in
reservoirs is crucial to the simulation of EGS. Discrete fracture network (DFN) as a discrete method enables the
explicit description of each fracture in the domain and has been employed in different scenarios of geothermal
extraction from fractured formations [33,34]. Gan and Elsworth [10] investigated the effect of the fracture
orientations, the stimulation strategies and the fracture density on the power generation in a 2D fractured reservoir
containing two sets of fractures and attempted to optimise the production strategy. Koh et al. [35] proposed a
numerical thermo-poroelastic reservoir model incorporating the fluid flow, temperature and fracture geomechanics.
Sun et al. [36] used COMSOL to implement the THM simulation of a highly fractured media based on a
local thermal non-equilibrium model. Yao et al. [37] then extended this approach to a three-dimensional version.
Many other THM models based on three-dimensional DFNs for geothermal development in EGS also have been
developed [11,25,38–41]. However, most of them are capable of modelling only single or multiple parallel fractures.
It is necessary to develop a robust three-dimensional model to effectively conduct coupling THM simulations of
large-scale discrete fracture networks in EGS.
This study proposes a strongly coupled thermo-hydro-mechanical model to simulate the complete process of the
geothermal extraction from EGS embedded with three-dimensional DFNs. The mechanical, hydraulic and thermal
sub-models in this coupling algorithm are formulated separately and iterated in a partitioned coupling way. Based
on a linear fracture model and a finite element method, the thermo-elastic solver is developed to model the aperture
variation of each fracture induced by the thermal expansion, over-pressure and external stresses. The fluid flow in
fracture networks is simulated by using a Darcy-flow-based unified pipe-network method (UPM). In addition, by
applying the local thermal non-equilibrium (LTNE) theory in the thermal solver, the process of the heat exchange
between fluid and rock matrix can be appropriately accounted for. This model is verified against an analytical
solution and the sensitivity analysis is conducted in respect of time step and mesh size. This model is further
applied to modelling and studying the 20-years geothermal productions of the Habanero EGS reservoir in Australia
under different injection and production strategies.
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 467

Fig. 1. Diagram of the strongly coupled process among the thermo-elastic solver, the fluid flow solver and the thermal solver.

2. Numerical models for strongly coupled THM analysis in EGS

In this section, the strongly coupled thermo-hydro-mechanical (THM) simulations in three-dimensional discrete
fracture networks (DFNs) are accomplished by properly formulating the interaction between different fields. The
complex interactive process between the thermal, mechanical and temperature fields is decomposed into three major
solvers: the thermo-elastic solver, the fluid flow solver and the thermal solver. The synchronisation of the important
parameters in the partitioned-coupling procedure is illustrated in Fig. 1. Additionally, some assumptions are made
to facilitate the simulation of the coupled THM process during the geothermal development in enhanced geothermal
systems (EGS),

(i) In hot dry rocks, there is no fluid existing in the rock matrix. Besides, the permeability of rock media is often
smaller than fractures by a few orders of magnitude. The discrete fracture network in the reservoir provides
the main flow pathway. It is reasonable to assume that the rock matrix is impervious and the seepage only
occurs inside fractures.
(ii) Water flow inside fractures is saturated, single-phase and incompressible. In the deep-depth formation, the
phase of water is unchanged with temperature because of the surrounding high pressure.
(iii) Local thermal non-equilibrium (LTNE) model is adopted to account for the heat exchange between the fluid
and the rock matrix, so that the fluid temperature might be different from the rock temperature on fracture
surfaces.
(iv) Hot dry rocks are assumed to be homogeneous, isotropic and linear elastic. Its mechanical properties are
independent of temperature.
(v) Stress wave is much faster than temperature conduction and flow convection so that the deformation of
the rock matrix can be solved by a static model, while the thermal conduction and flow convection are
transient.

The highly fractured rock embedded with a DFN can be discretised into adaptive conforming tetrahedral
elements by using the mesh techniques developed by Wang et al. [42]. The requirements of the meshes used in the
fluid/thermal and the mechanical simulations are different. The fluid/thermal simulations only demand the fractures
to be explicitly represented by the conforming triangles on fractures. Meanwhile, the mechanical simulations require
that the fracture interfaces are able to be separated physically in response to external stresses. Therefore, the
primarily generated mesh can be directly used in thermal/hydraulic simulations, but extra nodes and facets are
needed to be inserted into this mesh in order to guarantee the strongly discontinuous features of fractures in
mechanical simulations [43]. It is noted that only triangle elements on fractures are used in the hydraulic simulations
because the rock matrix is impervious.
468 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

Fig. 2. Illustration of the THM process near the fracture interface in a hot dry rock.

2.1. Thermo-elastic deformation of fractured rocks

In this section, the thermo-elastic deformation of a fractured rock is formulated and solved in terms of a
finite element method. As the fractures inside the rock mass are considered as strong discontinuities, the fracture
deformation in the THM process is basically caused by the external loads such as in situ stress, the thermal strain
induced by the heat loss in the hot dry rock and the overpressure induced by the fluid flow inside the fracture, as
shown in Fig. 2. According to the principle of virtual work, the equilibrium equation in a fractured rock mass Ω
can be expressed as
∫ ∫ ∫
σ : δεdΩ − t ∗ · δu dΓ − b · δu dΩ
Ω Γt Ω
∫
( )
+ T + t hydr o · JδuK dΓ = 0, (1)
Γd

where σ denotes the Cauchy stress tensor, b stands for the body force distribution, t ∗ specifies the prescribed traction
such as in-situ stress and fluid pressures on the boundary Γt , ε is the strain tensor, δε is the virtual strain field,
and δu is the virtual displacement field. T and F hydr o represent the contact traction vector and the force vector
contributed by the fluid pressure on fracture interfaces, respectively. JδuK is the virtual displacement jump over the
fracture interface. The total strain ε is given as
ε = εe + ε T (2)
]T
ε T = αT (T − T0 ) 1 ,
[
1 1 0 0 0 (3)
where εe and ε T denote the elastic part and thermal part of the strain, respectively. αT refers to the thermal expansion
coefficient for the rock matrix. T and T0 are the current temperature and the initial temperature of the rock matrix;
Furthermore, the traction over the fracture interface can be decomposed into the local coordinate system n-s-q. In
this study, a linear relationship between the normal displacement and the normal effective stress [44,45] is adopted to
model the fracture deformation, as shown in Fig. 3(a). The normal opening and the closure amount are constrained
by the maximum aperture and residual aperture, respectively. A similar model is also used on the shearing plane
s-q of the fracture interface as shown in Fig. 3(b). The decomposed tractions acting on Γd+ can be written as

n = kn nn JuK nn ,
T+ s = ks ns JuK ns ,
T
T+ T
T q+ = kq nqT JuK · nq (4)
JuK = u − u ,
− +
(5)
where k is the fracture stiffness. The subscripts n, s and q represent directions in the local coordinate system of
each triangle element pairs on the fracture interface. It is consisted of a unit normal vector nn outwards from Γd+
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 469

Fig. 3. Relations between stress and displacement of the fracture interface: (a) in the normal direction; (b) in the tangential direction.

and two unit tangential vectors ns and nq . u− and u+ are the displacements on both sides of the fracture interface.
All tractions acting on Γd− have the same magnitude as Γd+ , but opposite directions.
After substituting Eqs. (2)–(5) into Eq. (1), the governing equation can be written in a matrix form:

(K Ω + K n + K s + K q )u = F ext + F hydr o + F ther mal , (6)

in which
∫
KΩ = B T D B dΩ
Ω
∫
T T
K n = kn Ñ nn nnT Ñ dΓ
Γd
∫
T
K s = ks Ñ ns nsT Ñ dΓ
Γd
∫
T
K q = kq Ñ nq nqT Ñ dΓ (7)
Γd
∫ ∫
T ∗
F ext = N t dΓ + N T b dΩ
Γt Ω
∫
T
F hydr o = Ñ nn N̂ p dΓ
Γd
∫
F ther mal = B T Dε T dΩ ,
Ω
470 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

where K Ω refers to the global elastic stiffness of the rock matrix, K n , K s and K q are the global fracture stiffness
matrices contributed by the normal stiffness and the tangential stiffness, B stands for the strain–displacement matrix.
N, Ñ and N̂ are shape functions, D designates the matrix elastic matrix, F ext is the external force, F hydro denotes
the force imposed by the fluid pressure p inside fractures, F thermal represents the thermal force induced by the
temperature change of the rock matrix. The detailed derivation of Eqs. (6)–(7) is given in Appendix A. In this
study, one-point Gauss quadrature is used to approximate the stiffness matrices and force vectors in Eq. (7).
Adding to that, aperture distribution can be calculated when the displacement field is obtained after solving
Eq. (6). The total hydraulic aperture variation in fractures is dominated by the fracture opening, closure and shear
dilation, as shown below
wh = w0 + δn + wdilation , wr es ≤ wh ≤ wmax
wdilation = δt tan (φ) wdilation ≤ Ucs
δn = JuK · nn , δt = JuK · ns + JuK · nq , (8)
where wh is the hydraulic apertures of fractures used in the hydraulic simulation, and w0 is the initial aperture.
δn and δt denote the fracture opening and shear displacement, respectively. wmax and wr es are the maximum and
minimum fracture apertures, wdilation denotes the aperture promoted by shearing dilation, Φ represents the dilation
angle, Ucs is the maximum allowable dilation. In this study, the apertures of fracture tips are assumed to be equal
to the residual apertures and independent of the stress.

2.2. Saturated seepage in DFNs using UPM

The unified pipe-network method (UPM) was proposed by Ren et al. [46–48] to simulate the single-phase, multi-
phase, steady and transient flows in three-dimensional fractured and fractured-porous media. Following that, this
method is further applied to different projects involving DFNs, such as acidizing reaction, geothermal development,
nuclide migration, oil recovery and so on [25,49–56]. In this study, the unified pipe network is applied to modelling
the saturated flow through the DFN. The principle of UPM is briefly introduced in this section.
The mass conservation equation of the fluid flow in the DFN can be described as,
∂ φd ρ f
( )
+ ∇ · (ρv) = q, (9)
∂t
where ρ f refers to the fluid density, q denotes the source term, φd is the porosity of the fracture, and v represents
the flow velocity. According to Darcy’s law, flow velocity can be calculated as below,
k(
v = − ∇P − ρf g ,
)
(10)
µ
where k is the intrinsic permeability tensor for fractures, P denotes the fracture fluid pressure, µ is the dynamic
viscosity of the fluid, and g is gravitational acceleration.
In UPM, the edges of discretised triangles on the fracture surfaces are considered as a pipe network to characterise
the hydraulic behaviour of fractures. Fracture apertures are assigned at the vertices of each triangular element as
w2
shown in Fig. 4. According to the cubic law, the permeability at a vertex can be estimated by k = 12h , in which wh
is the aperture at this vertex. In this study, the fluid viscosity is temperature-dependent, so that the fluid viscosity at
the vertices can also be calculated using the nodal temperatures. In a triangular element ∆ijk , flow rate in the pipe
i j without gravity effect is calculated as,
qi j = ci j ( pi − p j ), (11)
where qi j refers to the flow rate outwards from nodes i to j through pipe i j, and ci j is the equivalent transmissivity
factor of the pipe ij and p is the fluid pressure at the vertex. The equivalent transmissivity can be estimated by
using the nodal transmissivity factors of the pipe endpoints as below
wh,i
3
bi j wh,
3
j bi j
ci = , cj =
6L i j µi 6L i j µ j
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 471

Fig. 4. Equivalent pipes of a triangle element on fractures in UPM.

ci c j bi j wh,i
3
wh,
3
j
ci j = = (12)
ci + c j 6L i j (µi wh,3
j + µ j wh,i )
3

where c specifies the transmissivity factor. wh and µ are aperture and fluid viscosity, respectively. Subscripts i and
j denote the nodes i and j, respectively. L i j is the pipe length, and bi j refers to the distance from the circumcentre
of the triangle to the midpoint of the pipe ij. The derivation of Eq. (12) is given in Appendix B.
The governing equation of the steady-state flow in each triangle element on fractures can be discretised into
⎡ ⎤⎡ ⎤ ⎡ ⎤
ci j + cik −ci j −cik pi qi
⎣ −ci j ci j + c jk −c jk ⎦ ⎣ p j ⎦ = ⎣q j ⎦ (13)
−cik −c jk cik + c jk pk qk
After the element transmissivity matrix are assembled into the global matrix, the global balance equation can be
written as,
Cp = Q (14)
where C denotes the global transmissivity matrix, p designates the fluid pressure on each node, and Q is the source
term. The fluid pressure field through the fracture network is obtained by solving Eq. (14).

2.3. Local thermal non-equilibrium (LTNE) for DFNs

The local thermal non-equilibrium [22] aims to describe the transient state in the process of the heat exchanging
between rock matrix and fluid before reaching the steady state, i.e. local thermal equilibrium. In this model, two
energy conservation equations are introduced to address the temperature difference between the fluid and the rock
skeleton. The first one is the transient balance equation for the fluid in fractures considering the thermal conduction,
thermal convection and heat exchange term, which can be expressed as,
∂Tf
φd ρ f c f + ρ f c f ∇ · v · T f − ∇ · φd λ f ∇T f = ah int Tm − T f ,
( ) ( ) ( )
(15)
∂t
where T f and Tm represent the temperature of fluid and rock matrix, respectively. c f is the heat capacity of fluid,
λ f is the thermal conductivity of fluid, a refers to the specific surface area, and h int is the heat transfer coefficient
between the rock matrix and the fluid.
As the rock matrix is dry and impervious, another transient balance equation for the rock matrix only includes
the thermal conduction and the heat exchange term. Hence, it can be written as below,
∂ Tm
(1 − φm )ρm cm − ∇ · ((1 − φm )λm ∇Tm ) = ah T f − Tm ,
( )
(16)
∂t
where φm is the porosity of the rock matrix, ρm is the rock density, cm is the heat capacity of the rock matrix, and
λm is the thermal conductivity of rock matrix.
Both Eqs. (15) and (16) can be discretised in the framework of UPM [50]. In this work, they are sorted and
assembled at the element level instead of as individual pipes for accelerating the computation. The discretised
472 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

balance equation for fluid phase at node i in a triangle element ∆ijk can be expressed as,
(t+∆t) (t)
T f,i − T f,i ∑ (
(t+∆t) (t+∆t)
)
φd ρ f c f Vi + cond
h in T f,i − T f,n
∆t n= j,k
∑ ( ) ( )
(t+∆t) (t+∆t) (t+∆t) (t+∆t)
+ h conv
f,in F T f,i , T f,n + 2h int Ai T f,i − Tm,i =0
n= j,k
Ain Φd λ f
h cond
f,in =
L in
(t+∆t)
ρ f c f qin
h conv
f,in = , (17)
L in
where Vi is the control volume for node i, Ain is the control area of the pipe in(n = j, k), Ai is the area of
the fracture interface governed by node i as shown in Fig. 4. The multiplier of 2 in front of Ai indicates the two
fracture walls, F is a function of Peclet number to determine which differencing scheme can be used properly in
the convection term [57,58]. In this study, the upwind differencing scheme is chosen for the convection in fractures.
Assuming the upwind nodes for pipe i j, ik and jk are i, k and j, respectively, then the assembled governing
equation for the triangle element is
[ ]
] T (t+∆t) (t)
f Tf ,
[ rate cond conv
h f + h f + h f + hint −hint f
= hrate (18)
T (t+∆t)
m
where
⎡ ⎤
Vi 0 0
φ ρ c
d f f ⎣
hrate
f = 0 Vj 0 ⎦
∆t 0 0 Vk
⎡ ⎤
h cond
f,i j + h f,ik
cond
−h condf,i j −h cond
f,ik
hcond
⎢ ⎥
cond cond
f =⎢⎣ −h f,i j h f,i j + h cond
f, jk −h cond
f, jk
⎥
⎦
−h cond
f,ik −h cond
f, jk h cond cond
f,ik + h f, jk
⎡ ⎤
conv conv
h f,i j 0 −h f,ik
hconv
⎢ ⎥
conv
f =⎢⎣−h f,i j h conv
f, jk 0 ⎥ ⎦
0 −h conv
f, jk h conv
⎡ ⎤ f,ik
Ai 0 0
hint = 2h int ⎣ 0 A j 0 ⎦ , (19)
0 0 Ak
where hcond
f and hconv
f are the element stiffness matrices of the thermal conduction and convection of the fluid in
fractures, respectively. hratef refers to the element matrix for the change rate of thermal energy of the fluid, hint is
[ ]T
the element stiffness matrix for energy transfer between the rock matrix and the fluid. T f = T f,i T f, j T f,k and
[ ]T
T m = Tm,i Tm, j Tm,k are the fluid and the rock temperatures at element vertices, respectively. It is noted that the
stiffness matrix of the thermal convection hconv f is non-symmetric.
Similarly, the discretised balance equation for the rock matrix at node i in a tetrahedral element i jkl can be
written as
(t+∆t) (t)
Tm,i − Tm,i ∑ (
(t+∆t)
)
(1 − φm )ρm cm Vi + Hincond Tm,i (t+∆t)
− Tm,n
∆t n= j,k,l
( )
(t+∆t) (t+∆t)
+h Tm,i − T f,i =0
Ain (1 − Φm )λm
cond
Hm,in = , (20)
L in
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 473

Then, the matrix form of the balance equation for the tetrahedral element can be expressed as
[ ]
(t+∆t)
T
m Tm ,
= hrate (t)
−hint hrate + hcond
[ ] f
m m + hint (21)
T (t+∆t)
m

where
⎡ ⎤
Vi 0 0 0
(1 − φ m )ρ f c f 0 V j 0 0⎥
hrate
⎢
m = ⎢
⎣ 0 0 Vk
⎥
∆t 0⎦
0 0 0 Vl
⎡ ∑ cond cond cond cond
⎤
Hm,in −Hm,i j −Hm,ik −Hm,il
⎢n= j,k,l ⎥
⎢ ∑ ⎥
⎢ −H cond cond cond cond
⎢ m,i j Hm, jn −Hm, jk −Hm, jl
⎥
⎥
cond n=i,k,l ⎥,
⎢ ⎥
hm = ⎢ ⎢
cond cond
∑
cond cond ⎥
(22)
⎢ −Hm,ik −Hm, jk Hm,kn −Hm,kl ⎥
⎢ n=i, j,l
⎥
⎢ ∑ ⎥
⎣ −H cond cond ⎦
m,il −H cond
m, jl
cond
−Hm,kl H m,ln
j=i, j,k

where hcond
m is the element stiffness matrix of the thermal conduction in the rock matrix; hrate
m is the element matrix
of the change rate of thermal energy in the rock matrix.
After assembling Eqs. (18) and (21), the global balance equation is
[ rate ] [ t+∆t ]
H f + H cond f + H conv
f + H int −H int Tf
(23)
−H int H rate
m + H cond
m + H int T t+∆t
m
[ ]
(t)
H rate T
= f f
,
(t)
H rate
m T m

where each node on the fracture surfaces has two degrees of freedom, namely the fluid temperature and the rock
temperature.
The fluid temperature can influence water properties, such as viscosity and density. However, water in fractures
is assumed not to be boiled because of the high-pressure environment in the deep-depth reservoir. Therefore, the
water density in a non-boiling system changes a little, but the viscosity decreased significantly with the increase in
fluid temperature up to 200 ◦ C [10,59]. An empirical formula is adopted to determine the temperature-dependent
viscosity of water [49,60]:
1
µ= , (24)
29.83(T f − 14.55)
where T f is the fluid temperature with units of degree Celsius.

2.4. Implementation of the strongly partitioned coupled THM model

The strongly partitioned coupling method for the THM model is established on basis of the interactions between
the different fields, as shown in Fig. 1. Fig. 5 further clarifies the iterative procedure in a time step for this coupled
model. The displacement field and fracture aperture distribution are firstly computed using the thermo-elastic solver.
The distribution of fracture fluid pressure is obtained after solving the fluid flow equation, and then the flow rate in
each pipe will be updated. According to this flow rate, the convection term in the thermal solver is also updated.
Afterwards, the temperatures of the fluid and the rock matrix are solved by the thermal solver. If the change
of the temperature, pressure and displacement are smaller than a specified tolerance, the iteration is convergent,
and it then moves to the next time step. Otherwise, the fluid viscosity, the hydro-induced force and the thermal-
induced force are updated and passed to the next iteration. As those solvers use different meshes to simulate, the
474 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

Fig. 5. Flow chart of implementing the strongly partitioned-coupling thermo-hydro-mechanical simulation.

data mapping between those meshes is essential and the method proposed by Ma et al. [43] is adopted in this
study.

3. Model verification
3.1. Thermal conduction and convection along a fracture based on local thermal non-equilibrium (LTNE) concept

In order to verify the coupled hydro-thermal (HT) model based on LTNE, a 3D model with a single fracture
is simulated and then compared with an analytical solution for 2D LTNE models derived by Zhao [61]. In
this verification, it is assumed that the thermal conduction in the rock only occurs in the vertical direction and
the conduction in the longitudinal direction is neglected. For verifying against the analytical solution, a large
thermal conduction coefficient of rock matrix is taken for minimising the effect of thermal conduction in the
rock matrix. As a result, the rock temperature on the fracture interface is almost the same as the outer boundary
temperature.
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 475

Fig. 6. Illustration of the three-dimensional thermal conduction and convection model.

Table 1
Model parameters for verifying LTNE model.
Parameter Symbol Value Unit
Fracture aperture wf 10 µm
Initial temperature T0 80 ◦C

Inlet temperature Tin 30 ◦C

Fluid density ρf 1000 kg/m3

Fluid viscosity µ 0.001 Pa s
Specific heat capacity of fluid cf 4200 J/(kg ◦ C)
Specific heat capacity of matrix cm 1000 J/(kg ◦ C)
Thermal conductivity of fluid λf 0.6 J/(m s ◦ C)
Thermal conductivity of matrix λm 1000 J/(m s ◦ C)
Heat transfer coefficient h int 10 W/(m2 ◦ C)

The analytical solution for the fluid temperature inside the fracture along the longitudinal direction can be
expressed as below,
T f (x) = T0 + (Tin − T0 )
⎡ ⎛ √( ⎞⎤
)2
x vρ c
f f vρ c
f f 16hλ m ⎠⎦ ,
× exp ⎣ ⎝ − + (25)
2 λf λf 2wh λ f (λm + h H )

where v is the flow velocity in the x direction and H is the height of the domain.
As shown in Fig. 6, the verified model has the dimension of 1 m x 0.25 m x 0.25 m and contains a horizontal
fracture in the middle of the domain. The rock matrix is impervious with an initial temperature Tm,0 = 80 ◦ C.
The top and bottom surfaces maintain a constant temperature of Tm,0 , while water with an initial temperature
T f,0 = 20 ◦ C is injected at a constant velocity from the left side. The fluid temperature at the outlet is unknown in
prior, so that outlet boundary is set as thermal outflow boundary. The domain is discretised into 115,813 tetrahedral
elements. All input parameters of this simulation are listed in Table 1.
The numerical and analytical solutions in different flow velocities are compared in Fig. 7. It is seen that the
temperature at the outlet decreases with the faster injection flow. The numerical results for different flow velocities
stay consistent with the analytical solution. Therefore, it is verified that the LTNE-based UPM method is capable
of accurately simulating thermal convection and conduction inside the fracture.

3.2. Sensitivity analysis for the strongly coupled THM simulation

In this section, the simulation in an HDR mass containing an elliptical fracture is implemented with different
time steps and mesh sizes for validating the numerical sensitivity and stability of the proposed method. As shown
in Fig. 8, the dimension of the domain is 10 m × 10 m × 5 m. The lengths of semi-major and semi-minor axes
of the elliptical fracture are 4 m and 3 m, respectively. Initial temperature of the rock matrix and the fluid inside
the fracture is 100 ◦ C, while the rock bottom is fixed at the same temperature. 20 ◦ C cold water is continuously
injected into the hot rock mass and extracted at the outlet through the fracture. All model parameters adopted for
476 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

Fig. 7. Fluid temperature distributions along the fracture using different flow velocities.

Fig. 8. Geometry and boundary conditions of the model to verify the THM interaction during the thermal extraction.

this verification are listed in Table 2. The fluid temperature along line A–B on the fracture will be extracted and
compared.
The fluid temperature distributions along A–B for different time steps and mesh sizes are shown in Fig. 9. As
Fig. 9(a) shows, the fluid temperature at t = 2 days is sensitive to the selection of the time step. In the simulation
using the time step of 2 days, only one step is computed, so that its error is relatively larger than others. However,
the result converges as the time step decreases. The effect of time step reduces with increasing simulated time.
The results at t = 50 days using different time steps converge to the same line because all of them have enough
computation steps to reach a stable result. Therefore, the strategy of adaptive time step is suitable for long-term
simulation for keeping the accuracy and reducing time cost. Fig. 9(b) shows the fluid temperatures along AB at
t = 10 days using different element sizes. The solution is convergent with the number of elements increases.
However, the solution nearby the fracture tips is more sensitive to element size. This is because that the element
size around fracture tips can influence the scope of the fracture tips, on which the nodal apertures are fixed and
independent of the thermal and pressure fields. As the real fracture tip zone is an extremely small region, smaller
element sizes around fracture tips are suggested. In these simulations, after the number of elements is greater than
101,064, the effect of fracture tips will become minor. The simulation results of fracture aperture, fluid temperature,
rock temperature and pressure distribution on the 10th day in the sample of ∆t = 0.5 day and 101,064 elements
are displayed in Fig. 10. LTNE model is capable of capturing the different temperatures of rock and water on the
fracture interface. The fracture aperture will also be altered from the constant state due to the thermal shrinking of
rock medium and fluid overpressure inside the fracture.
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 477

Fig. 9. Fluid temperature distributions along A–B using (a) different time steps; (b) and different mesh sizes.

Fig. 10. Simulation results on the fracture after 10 days: (a) water temperature inside the fracture; (b) rock temperature on the fracture top
interface; (c) fluid pressure distribution; (d) hydraulic aperture distribution.
478 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

Table 2
Model parameters for the strongly coupled THM simulation.
Symbol Value Unit
Rock properties
Young’s modulus E 10 GPa
Poisson’s ratio ν 0.25
Rock density ρm 2700 kg/m3
Rock thermal expansion coefficient αT 2 × 10−6 1/◦ C
Thermal conductivity of rock λm 3.5 J/(m s ◦ C)
Specific heat capacity of rock cm 790 J/(kg ◦ C)
Fluid properties
Fluid density ρf 1000 kg/m3
Fluid viscosity µ 0.001 Pa s
Thermal conductivity of fluid λf 0.6 J/(m s ◦ C)
Specific heat capacity of fluid cf 4200 J/(kg ◦ C)
Fracture properties
Initial fracture aperture w0 100 µm
Normal fracture stiffness kn 10 GPa/m2
Tangential fracture stiffness kt 1 GPa/m2
Dilation angle φ 10 degrees
Critical shear displacement for dilation Ucs 1 mm
Others
Initial fluid temperature in fractures T f,0 100 ◦C

Initial rock temperature Tm,0 100 ◦C

Inlet fluid temperature Tin 20 ◦C

Heat transfer coefficient h int 15 W/(m2 ◦ C)

Injection pressure Pin 0.5 MPa
Outlet pressure Pout 0.1 MPa

4. Habanero enhanced geothermal reservoir

4.1. Project background

Habanero enhanced geothermal reservoir has been developed by Geodynamics Limited in Cooper Basin of South
Australia since 2002, and the first private commercial EGS plant for electricity production was online in 2013 [3].
The basement rock in the reservoir is water-saturated fractured granite that retains temperature up to 248.3 ◦ C at a
depth of 4391 m and is tightly confined by overlying sedimentary formations [62]. The first well Habanero 1 (H1)
was drilled 753 m into the basement rock in 2002. It was primarily stimulated in 2003 resulting in over 28,000
seismic events and then the subsequent re-stimulation in 2005 produced 16,000 more seismic events and expanded
the scope of the fractured zone [63]. Habanero 2 (H2) and Habanero 3 (H3) are another two wells drilled at 500 m
to the south-west and 600 m to north-east of H1, respectively. Nevertheless, H2 lost its function in 2005 due to a
blockage. Habanero 4 (H4) was completed as a production well in 2013, and it is used as a pilot plant currently. The
depletion and open-loop circulation tests between H1 and H3 were carried out in 2008 and proved that a hydraulic
connection between H1 and H3 exists [64].
Many approaches have been developed to numerically simulate and quantify the process of thermal production
in Habanero project. Llanos et al. [65] developed and calibrated a TOUGH2 model to simulate thermal productions
using different injection rates for optimising the well layouts. Xu et al. [66] proposed a simplified model to simulate
the coupled hydro-thermal process between H1 and H3 in a three-dimensional conditional fracture model. However,
simplifying fracture network into a connectivity path will miss the details on fractures and possibly underestimate
the thermal exchange between the fluid and the rock mass. Besides, the stress reaction induced by fluid pressure
and thermal expansion is also a significant factor in the thermal extraction progress. Sun et al. [36] proposed a
two-dimensional coupled THM model based on the local thermal non-equilibrium theory to simulate the thermal
extraction process between two wells in Habanero reservoir including a stochastically generated fracture network. In
this section, a three-dimensional complex discrete fracture network (DFN) is applied to characterise the significant
hydraulic connectivity between H1 and H3 wells. Based on the verified THM model, the closed-loop thermal
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 479

Fig. 11. Illustration of the closed-loop circulation model for H1–H3 doublet in Habanero reservoir.

production in the doublet (H1–H3) for a production period of 20 years is simulated. Moreover, the effects of injection
pressure and production pressure on production efficiency will be investigated and discussed.

4.2. Generation and discretisation of the EGS

As shown in Fig. 11, a hot fractured granite mass with a dimension of 800 m × 800 m × 300 m is considered.
Injection well H1 and production well H3 are located on the south-west corner and north-east corner, respectively.
All boundaries are assumed to be impervious for modelling closed-loop production. The granite mass contains
300 polygonal fractures, which are uniformly positioned in the domain. Orientations and sizes of the fractures are
generated according to Fisher’s distribution and lognormal distribution, respectively. Referring to the data in [67],
Fisher’s coefficient k = 1, mean fracture size µ = 80 m and variance σ 2 = 12 000 m2 are adopted. The fracture
size threshold of 50 m and 200 m are chosen for filtering the extremely small or large fractures in the domain.
Subsequently, this model is primarily discretised into 656,712 conforming tetrahedral elements and 113 898 nodes.
The original mesh in thermal simulation and the mesh in the hydraulic simulation are displayed in Fig. 12. In
the thermal simulation, there are 227,796 degrees of freedom (DOFs) in total because each node has two DOFs
including rock temperature and fluid temperature. In the hydraulic simulation, only pressure nodes on the fractures
are considered, so that the total number of DOFs is only 87,429. After adding extra nodes and facets into the
original mesh for modelling strong discontinuities in the thermo-elastic simulation, the total number of DOFs for
the simulation of the fractured reservoir deformation is 627,123. In terms of DOFs, the thermos-elastic solver has
the most considerable computational intensity.

4.3. Boundary conditions and model parameters

Cold water is circulating from H1 and then though fractures to H3. The fractures are saturated with initial
water pressure of 35 MPa before water is injected [66]. Water will be injected at H1 with constant pressure PH 1
and fixed temperature Tin , and then flow through the fractures to be extracted from H2 with constant operating
pressure PH 2 . In this process, if the current fluid pressure in the fracture is smaller than the initial water pressure,
the effective pressure will be imposed inward on the fracture, so that the fracture will tend to close. In contrast,
if the fluid pressure is larger than the initial water pressure, the overpressure will drive the fracture to open. The
20-years geothermal production using different injection and production pressures are simulated and evaluated. Input
parameters in these simulations are listed in Table 3 and for other material parameters that are not mentioned in
this section, refer to Table 2.

4.4. Simulation results and discussion

The comprehensive understanding of the response of the reservoir under different injection and production
pressures are vital to the optimisation of geothermal extraction. Four cases are simulated, and then the results
480 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

Fig. 12. (a) Original conforming tetrahedral mesh for the thermal simulation; (b) DFN mesh for the hydraulic simulation.

Table 3
Model parameters for simulation of H1–H3 doublet in Habanero project.
Symbol Value Unit
Fracture properties
Initial fracture aperture w0 100 µm
Maximum aperture wmax 400 µm
Residual aperture wr es 10 µm
Normal fracture stiffness kn 50 GPa/m2
Tangential fracture stiffness kt 10 GPa/m2
Others
Initial fluid temperature in fractures T f,0 250 ◦C

Initial rock temperature Tm,0 250 ◦C

Inlet fluid Temperature Tin 20 ◦C

Heat transfer coefficient h int 1000 W/(m2 ◦ C)

Initial water pressure in reservoir P0 30 MPa
Case 1
Injection pressure PH 1 40 MPa
Production pressure PH 2 34 MPa
Case 2
Injection pressure PH 1 50 MPa
Production pressure PH 2 44 MPa
Case 3
Injection pressure PH 1 44 MPa
Production pressure PH 2 34 MPa
Case 4
Injection pressure PH 1 42 MPa
Production pressure PH 2 32 MPa

for different injection and production pressures are displayed in Fig. 13. The pressure drops in both case 1 and case
2 are the same, i.e. 6 MPa, while case 3 and case 4 have higher pressure drops of 10 MPa. In addition, case 2 has
the highest production pressure which is higher than the initial fluid pressure by 9 MPa. The production pressures of
both cases 1 and 3 are lower than the initial fluid pressure by 1 MPa and case 4 has the lowest production pressure
of 32 MPa. Although the pressure difference in case 2 is the lowest, its average operating pressure is the highest, 47
MPa. Besides, the average pressure for cases 1 and 4 is 37 MPa, which is slightly smaller than 39 MPa in case 3.
Evolutions of the average outlet temperature for those four cases are shown in Fig. 13(a). The average water
temperatures at the production well for all cases are stable at 250 ◦ C in the first two years. In due course, the
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 481

Fig. 13. Simulation results in the production period of 20 years: (a) average outlet temperatures; (b) average reservoir temperatures; (c)
reservoir power generations; (d) average fracture apertures.

temperature in case 2 began to decline and is reduced by 30.20%, giving 173.50 ◦ C in 20 years. The production
temperature in case 3 reduced gradually after 6 years and eventually reached 219.34 ◦ C with a drop of 12.26%. A
stable heat production trend was observed in both cases 1 and 4, in which the water temperatures declined by only
0.27 ◦ C and 3.81 ◦ C, respectively. As shown in Fig. 13(b), the variations of average rock temperature are evident in
all cases. The geothermal depletion of the rock reservoir in case 2 is the most distinct and the fastest, which might
be caused by the lowest average water temperature and most rapid decline in the outlet temperature. In cases 1 and
4, although the outlet temperature drops are slight, the reductions in the average rock temperature are remarkable.
Moreover, the geothermal depletion rates in all cases are becoming flat with time.
The power generation from the reservoir against time is displayed in Fig. 13(c). Although the power generation
in case 2 is the highest, it lost 41.7% of the peak performance after 20 years. In contrast, case 3 only lost 19.94% of
its efficiency. However, both cases 1 and 4 are able to maintain a stable power generation for 20 years. The trends of
power generation are different with different average outlet temperatures. This might be caused by the fact that the
flow rate at the production well changes with time due to the temperature-dependent viscosity and aperture variation.
Fig. 13(d) shows the aperture variation in this period. The mean fracture aperture in the reservoir increases steadily
with time from an initial value that is dominated by the operating pressures at two wells. The steady rise of the
mean aperture is contributed by the thermal shrinking of the rock matrix due to the temperature drop. In this DFN
model, the magnitudes of the injection well pressure and the mean operation pressure together dominate the average
fracture apertures. In case 2, both of injection and production pressures are above the initial fluid pressure, so that
the overpressures across these fractures drive these fractures to open. Although both cases 1 and 4 have the same
482 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

Fig. 14. Contours of the water temperature in case 3 after: (a). 1 year; (b) 5 years; (c) 10 years; (d) 20 years.

mean operating pressure, the larger injection pressure in case 1 drives the fractures to open more. Besides, the
aperture variations for all cases maintain the same trend throughout 20 years.
Contours of the fluid temperature in the DFN model and sliced contours of the rock temperature in the reservoir
for case 3 at different times are shown in Figs. 14–16, respectively. It is easy to see that the cooling process in
the model is non-uniform. The low-temperature zones for both water and rock matrix preferentially expand to the
south-east and north-west regions and then return toward the production well. These cooling regions mainly surround
the preferential flow paths that are dominated by the fracture connectivity and the interaction among stress field,
thermal field and fluid pressure field. These simulations quantify and visualise the complicated evolution of the
temperature fields along with time.

5. Conclusion
A strongly coupled THM model developed in this study approaches the simulation of the three-dimensional
geothermal production process in the enhanced geothermal system containing discrete fracture networks. Based
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 483

Fig. 15. Contours of the sliced rock temperature in case 3 after: (a) 1 year; (b) 5 years; (c) 10 years; (d) 20 years.

Fig. 16. Evolutions of the sliced rock temperatures in case 3 from different views.
484 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

on the linear fracture model and the thermo-elastic finite element method, the variation of fracture apertures in
response to thermal expansion, fluid overpressure and external stresses can be appropriately simulated. The local
thermal non-equilibrium theory separates the rock temperature and water temperature. Moreover, this proposed
approach is verified by comparing with the analytical solution and the sensitivity analysis of time steps and mesh
size have been carried out.
Using this proposed THM model, the closed-loop geothermal production of H1–H3 wells in Habanero project for
20 years is simulated and discussed. A complex three-dimensional stochastically generated discrete fracture network
is adopted to characterise the hydraulic connectivity between H1 and H3. According to the numerical results, the
generation power, the depletion of the reservoir and the production efficiency in this model are highly correlated to
the injection and production pressures. Keeping both pressures higher than the initial fluid pressure in fractures is
capable of boosting power generation significantly, but the depletion of geothermal energy and the efficiency losses
are apparent and fast. By appropriately controlling the operating pressures at both wells, it is possible to constantly
and stably supply power at specified levels during the whole period.
The verification and the case study have demonstrated that the proposed method is an efficient, effective and
robust way to model the strongly coupled THM process in three-dimensional large-scale EGS. In future works,
the saturated–unsaturated flow model accounting for the leakage effect in fractured porous reservoirs, and different
working fluid model such as supercritical carbon dioxide model in geothermal extraction will be extended and
developed on the basis of the developed method.

Acknowledgement
The authors would like to greatly acknowledge the Research Training Program (RTP) awarded by School of
Civil, Environmental and Mining Engineering, the University of Western Australia.

Appendix A. Derivation of the governing equation of the deformation of fractured rocks

The governing equation of a 3D fractured rock mass with 2D fractures is given in Eq. (1), and it can be rewritten
as
G (u, δu) + G d (u, δu) = 0 (A.1)
∫ ∫ ∫
G (u, δu) = σ : δε dΩ − t · δu dΓ −
∗
b · δu dΩ (A.2)
∫Ω Γt Ω

G d (u, δu) =
( )
T − t hydr o · JδuK dΓ = 0, (A.3)
Γd

By introducing shape function N, the displacement field u e (x) inside a finite element can be interpolated as
ue (x) = N u e , (A.4)
]T
ue (x) = u v w
[
(A.5)
]T
u e = u 1 v1 w 1 u 2 v2 w 2 . . .
[
(A.6)
⎡ ⎤
N1 0 0 N2 0 0 ···
N = ⎣ 0 N1 0 0 N 2 0 · · ·⎦ , (A.7)
0 0 N1 0 0 N2 · · ·
where u, v, w are displacement components at a node.
The strain vector ε is given as
]T
ε = εx ε y εz γx y γ yz γzx
[

∂u ∂v ∂w ∂v ∂u ∂w ∂v ∂w ∂u T
[ ]
= + + +
∂ x ∂ y ∂z ∂ x ∂y ∂y ∂z ∂ x ∂z (A.8)
⎡ ⎤
u
= L ⎣v ⎦ ,
w
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 485

where L is a differentiable operator and is given as

∂
⎡ ⎤
⎢∂x 0 0 ⎥
⎢
⎢ 0 ∂ ⎥
0⎥
∂y
⎢ ⎥
∂
⎢ ⎥
⎢ ⎥
⎢ 0 0 ⎥
L=⎢ ∂
⎢ ∂z ⎥,
⎥
(A.9)
∂
⎢
⎥
0⎥
⎢∂x ∂y
⎢
⎥
⎢ ∂ ∂ ⎥
⎢ 0
⎢ ⎥
∂ y ∂z ⎥
⎥
⎢
⎣ ∂ ∂ ⎦
0
∂x ∂z
Substituting the interpolated displacement field into Eq. (A.8), the strain field can be written as
ε = L N u = Bu, (A.10)
where B is strain–displacement matrix.
According to Hook’s law for an elastic body, the stress vector is
σ = Eεe , (A.11)
Substituting the elastic part of strains εe from Eq. (2) into Eq. (A.11):
σ = D (ε − ε T ) = D Bq − Dε T , (A.12)
Next, substituting Eqs. (A.4), (A.10) and (A.12) into Eq. (A.2):
∫ ∫ ∫
G (u, δu) = δu B D Bu dΩ −
T T
δu B Dε T dΩ −
T T
δu T N T t ∗ dΓ
Ω∫ Ω Γt (A.13)
− δu T N T b dΩ ,
Ω
Then, the shape functions are also adopted to interpolate the relative displacement Jue (x)K over triangle elements
of the fracture interface,
[ ]
e−
e+ u
Jue (x)K = N(ue− = Ñ u e
[ ] d
d − ud ) = N −N (A.14)
ude+
gne = nnT Jue K = nnT Ñ u e (A.15)
gse = nsT e
Ju K = nsT Ñ u e (A.16)
gqe = nqT e
Ju K = nqT Ñ u e , (A.17)
where ue−
dand ude+
denote the displacement of each node in a triangular element on and Γd− Γd+ ,
respectively.
According to Eq. (4), the tractions acting on the triangle pair of the fracture interface can be calculated:
T e = kn gne nn + ks gne ns + kq gne nq , (A.18)
Substituting Eqs. (A.15)–(A.17) into Eq. (A.18),
T e = kn nn nnT Ñ u e + ks ns nsT Ñ u e + kq nq nqT Ñ u e , (A.19)
The next step is to introduce the shape function N̂ to obtain the pressure field P e in the hydraulic triangular element
of the fracture interface,
P e = N̂ pe (A.20)
e T
p = [ p1 p2 p3 ] (A.21)
N̂ = [N1 N2 N3 ] , (A.22)
486 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

where p1 , p2 and p3 are the pressure at each vertex. Note that the hydraulic pressures at nodes are calculated by
the hydraulic solver using the triangular elements of the fractures.
The hydraulic traction can be obtained by multiplying the pressure and the associated unit normal vector,
t hydr o = nn P e = nn N̂ pe , (A.23)
Substituting Eqs. (A.14), (A.19) and (A.23) into Eq. (A.3),
∫
G d (u, δu) =
( )
T − t hydr o · JδuK dΓ
Γd
∫ ∫
= T · JδuK dΓ − t hydr o · JδuK dΓ
Γd Γd
∫
T (
δu T Ñ kn nn nnT Ñ + ks ns nsT Ñ + kq nq nqT Ñ u dΓ
)
=
Γd
∫
T
− δu T Ñ nn N̂ p dΓ , (A.24)
Γd

Substituting Eqs. (A.13) and (A.24) into (A.1) and simplifying,

(∫ ∫ ∫
T T T T T
B D B dΩ +kn Ñ nn nn Ñ dΓ + ks Ñ ns nsT Ñ dΓ
Ω
∫Γd ) Γd
T T
+ kq Ñ nq nq Ñ dΓ u
Γd (A.25)
∫ ∫ ∫
T
= N T t ∗ dΓ + N T b dΩ + Ñ nn N̂ p dΓ
Γt Ω Γd
∫
+ B T Dε T dΩ ,
Ω
Let
∫
KΩ = B T D B dΩ
Ω∫
T T
K n = kn Ñ nn nnT Ñ dΓ
∫ Γd
T
K s = ks Ñ ns nsT Ñ dΓ
∫Γd
T
K q = kq Ñ nq nqT Ñ dΓ (A.26)
Γd
∫ ∫
F ext = N T t ∗ dΓ + N T b dΩ
Γt Ω
∫
T
F hydr o = Ñ nn N̂ p dΓ
Γd
∫
F ther mal = B T Dε T dΩ ,
Ω
Then, the governing equation (A.25) has the following form:
(K Ω + K n + K s + K q )u = F ext + F hydr o + F ther mal . (A.27)

Appendix B. Derivation of the equivalent transmissivity of a pipe with various apertures

A pipe ij is given and the hydraulic apertures at both endpoints are wh,i and wh, j , respectively. Assuming the
middle point of the pipe is c and each endpoint governs the flow of the half pipe, then the flow rates in both half
pipes can be calculated according to Eq. (11)
qic = ci ( pi − pc ) (B.1)
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 487

qcj = c j ( pc − p j ) (B.2)
where ci and cj specify the transmissivity factors in the half pipes governed by nodes i and j, respectively. p denotes
the nodal pressure, and q refers to the flow rate of the pipe.
Assuming the equivalent transmissivity factor of the entire pipe is ci j , the flow rate in the pipe is calculated,
qi j = ci j ( pi − p j ) (B.3)
qi j = ci j (( pi − pc ) + ( pc − p j )) (B.4)
Substitute Eqs. (B.1)–(B.2) into (B.4),
qi j qic qcj
= + (B.5)
ci j ci cj
According to the conservation of mass, qi j = qic = qcj , and then Eq. (B.4) can be simplified,
1 1 1
= + (B.6)
ci j ci cj
ci c j
ci j = (B.7)
ci + c j

References
[1] J.W. Lund, D.H. Freeston, T.L. Boyd, Direct utilization of geothermal energy 2010 worldwide review, Geothermics 40 (2011) 159–180,
http://dx.doi.org/10.1016/j.geothermics.2011.07.004.
[2] J.W. Lund, T.L. Boyd, Direct utilization of geothermal energy 2015 worldwide review, Geothermics 60 (2016) 66–93, http://dx.doi.
org/10.1016/j.geothermics.2015.11.004.
[3] P. Olasolo, M.C. Juárez, M.P. Morales, S.D. Ámico, I.A. Liarte, Enhanced geothermal systems (EGS): A review, Renew. Sustain.
Energy Rev. 56 (2016) 133–144, http://dx.doi.org/10.1016/j.rser.2015.11.031.
[4] S. Kelkar, G. WoldeGabriel, K. Rehfeldt, Lessons learned from the pioneering hot dry rock project at Fenton Hill, USA, Geothermics
63 (2016) 5–14, http://dx.doi.org/10.1016/j.geothermics.2015.08.008.
[5] D. Duchane, D. Brown, Hot Dry Rock Geothermal Energy Development in the USA, Los Alamos Natl. Lab., Los Alamos, NM, 2000.
[6] T. Kohl, T. Mégel, Predictive modeling of reservoir response to hydraulic stimulations at the European EGS site Soultz-sous-Forêts,
Int. J. Rock Mech. Min. Sci. 44 (2007) 1118–1131, http://dx.doi.org/10.1016/j.ijrmms.2007.07.022.
[7] G. Zimmermann, A. Reinicke, Hydraulic stimulation of a deep sandstone reservoir to develop an enhanced geothermal system:
Laboratory and field experiments, Geothermics 39 (2010) 70–77, http://dx.doi.org/10.1016/j.geothermics.2009.12.003.
[8] M. AbuAisha, B. Loret, D. Eaton, Enhanced geothermal systems (EGS): Hydraulic fracturing in a thermo-poroelastic framework, J.
Pet. Sci. Eng. 146 (2016) 1179–1191, http://dx.doi.org/10.1016/j.petrol.2016.07.027.
[9] M.W. McClure, R.N. Horne, An investigation of stimulation mechanisms in enhanced geothermal systems, Int. J. Rock Mech. Min.
Sci. 72 (2014) 242–260, http://dx.doi.org/10.1016/j.ijrmms.2014.07.011.
[10] Q. Gan, D. Elsworth, Production optimization in fractured geothermal reservoirs by coupled discrete fracture network modeling,
Geothermics 62 (2016) 131–142, http://dx.doi.org/10.1016/j.geothermics.2016.04.009.
[11] Y. Zhao, Z. Feng, Z. Feng, D. Yang, W. Liang, THM (THermo-hydro-mechanical) coupled mathematical model of fractured media
and numerical simulation of a 3D enhanced geothermal system at 573 K and buried depth 6000–7000 M, Energy 82 (2015) 193–205,
http://dx.doi.org/10.1016/j.energy.2015.01.030.
[12] A.R. Shaik, S.S. Rahman, N.H. Tran, T. Tran, Numerical simulation of fluid-rock coupling heat transfer in naturally fractured geothermal
system, Appl. Therm. Eng. 31 (2011) 1600–1606, http://dx.doi.org/10.1016/j.applthermaleng.2011.01.038.
[13] T. Guo, F. Gong, X. Wang, Q. Lin, Z. Qu, W. Zhang, Performance of enhanced geothermal system (EGS) in fractured geothermal
reservoirs with CO2 as working fluid, Appl. Therm. Eng. 152 (2019) 215–230, http://dx.doi.org/10.1016/j.applthermaleng.2019.02.024.
[14] B. Faigle, M.A. Elfeel, R. Helmig, B. Becker, B. Flemisch, S. Geiger, Multi-physics modeling of non-isothermal compositional flow
on adaptive grids, Comput. Methods Appl. Mech. Engrg. 292 (2015) 16–34.
[15] F. Jiang, L. Luo, J. Chen, A novel three-dimensional transient model for subsurface heat exchange in enhanced geothermal systems,
Int. Commun. Heat Mass Transf. 41 (2013) 57–62, http://dx.doi.org/10.1016/j.icheatmasstransfer.2012.11.003.
[16] F. Jiang, J. Chen, W. Huang, L. Luo, A three-dimensional transient model for EGS subsurface thermo-hydraulic process, Energy 72
(2014) 300–310, http://dx.doi.org/10.1016/j.energy.2014.05.038.
[17] W. Cao, W. Huang, F. Jiang, A novel thermal–hydraulic–mechanical model for the enhanced geothermal system heat extraction, Int.
J. Heat Mass Transfer 100 (2016) 661–671, http://dx.doi.org/10.1016/j.ijheatmasstransfer.2016.04.078.
[18] Y.-C. Zeng, Z. Su, N.-Y. Wu, Numerical simulation of heat production potential from hot dry rock by water circulating through two
horizontal wells at Desert Peak geothermal field, Energy 56 (2013) 92–107, http://dx.doi.org/10.1016/j.energy.2013.04.055.
[19] B. Wu, G. Zhang, X. Zhang, R.G. Jeffrey, J. Kear, T. Zhao, Semi-analytical model for a geothermal system considering the effect of
areal flow between dipole wells on heat extraction, Energy 138 (2017) 290–305, http://dx.doi.org/10.1016/j.energy.2017.07.043.
[20] A. Borgia, K. Pruess, T.J. Kneafsey, C.M. Oldenburg, L. Pan, Numerical simulation of salt precipitation in the fractures of a
CO2-enhanced geothermal system, Geothermics 44 (2012) 13–22, http://dx.doi.org/10.1016/j.geothermics.2012.06.002.
488 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489

[21] G.A. Zyvoloski, B.A. Robinson, H.S. Viswanathan, Generalized dual porosity: A numerical method for representing spatially variable
sub-grid scale processes, Adv. Water Resour. 31 (2008) 535–544, http://dx.doi.org/10.1016/j.advwatres.2007.11.006.
[22] R. Gelet, B. Loret, N. Khalili, A thermo-hydro-mechanical coupled model in local thermal non-equilibrium for fractured HDR reservoir
with double porosity, J. Geophys. Res. Solid Earth 117 (2012) http://dx.doi.org/10.1029/2012JB009161, n/a-n/a.
[23] J. Taron, D. Elsworth, Thermal–hydrologic–mechanical–chemical processes in the evolution of engineered geothermal reservoirs, Int.
J. Rock Mech. Min. Sci. 46 (2009) 855–864, http://dx.doi.org/10.1016/j.ijrmms.2009.01.007.
[24] L.H. Odsæter, T. Kvamsdal, M.G. Larson, A simple embedded discrete fracture–matrix model for a coupled flow and transport problem
in porous media, Comput. Methods Appl. Mech. Engrg. 343 (2019) 572–601.
[25] Y. Wang, T. Li, Y. Chen, G. Ma, Numerical analysis of heat mining and geological carbon sequestration in supercritical CO2 circulating
enhanced geothermal systems inlayed with complex discrete fracture networks, Energy (2019) http://dx.doi.org/10.1016/j.energy.2019.
02.055.
[26] M. Nejati, A. Paluszny, R.W. Zimmerman, A finite element framework for modeling internal frictional contact in three-dimensional
fractured media using unstructured tetrahedral meshes, Comput. Methods Appl. Mech. Engrg. 306 (2016) 123–150, http://dx.doi.org/
10.1016/j.cma.2016.03.028.
[27] L.F. FAN, X.W. Yi, G.W. Ma, Numerical manifold method (NMM) simulation of stress wave propagation through fractured rock mass,
Int. J. Appl. Mech. 05 (2013) 1350022, http://dx.doi.org/10.1142/S1758825113500221.
[28] X. Zhou, L. Fan, Z. Wu, Effects of microfracture on wave propagation through rock mass, Int. J. Geomech. 17 (2017) 04017072,
http://dx.doi.org/10.1061/(ASCE)GM.1943-5622.0000947.
[29] Z. Wu, L. Fan, Q. Liu, G. Ma, Micro-mechanical modeling of the macro-mechanical response and fracture behavior of rock using the
numerical manifold method, Eng. Geol. 225 (2017) 49–60, http://dx.doi.org/10.1016/j.enggeo.2016.08.018.
[30] G.W. Ma, H.D. Wang, L.F. Fan, B. Wang, Simulation of two-phase flow in horizontal fracture networks with numerical manifold
method, Adv. Water Resour. 108 (2017) 293–309, http://dx.doi.org/10.1016/j.advwatres.2017.08.013.
[31] Y. Zhang, R. Lackner, M. Zeiml, HA. Mang, Strong discontinuity embedded approach with standard SOS formulation: Element
formulation, energy-based crack-tracking strategy, and validations, Comput. Methods Appl. Mech. Eng. 287 (2015) 335–366, http:
//dx.doi.org/10.1016/j.cma.2015.02.001.
[32] Y. Zhang, X. Zhuang, Cracking elements : A self-propagating strong discontinuity embedded approach for quasi-brittle fracture, Finite
Elem. Anal. Des. 144 (2018) 84–100, http://dx.doi.org/10.1016/j.finel.2017.10.007.
[33] P. Fu, S.M. Johnson, Y. Hao, C.R. Carrigan, Fully coupled geomechanics and discrete flow network modeling of hydraulic fracturing
for geothermal applications, in: 36th Stanford Geotherm. Work., 2011.
[34] H. Hofmann, T. Babadagli, G. Zimmermann, Hot water generation for oil sands processing from enhanced geothermal systems: Process
simulation for different hydraulic fracturing scenarios, Appl. Energy 113 (2014) 524–547, http://dx.doi.org/10.1016/j.apenergy.2013.07.
060.
[35] J. Koh, H. Roshan, S.S. Rahman, A numerical study on the long term thermo-poroelastic effects of cold water injection into naturally
fractured geothermal reservoirs, Comput. Geotech. 38 (2011) 669–682, http://dx.doi.org/10.1016/j.compgeo.2011.03.007.
[36] Z. Sun, X. Zhang, Y. Xu, J. Yao, H. Wang, S. Lv, Z. Sun, Y. Huang, M. Cai, X. Huang, Numerical simulation of the heat
extraction in EGS with thermal-hydraulic-mechanical coupling method based on discrete fractures model, Energy 120 (2017) 20–33,
http://dx.doi.org/10.1016/j.energy.2016.10.046.
[37] J. Yao, X. Zhang, Z. Sun, Z. Huang, J. Liu, Y. Li, Y. Xin, X. Yan, W. Liu, Numerical simulation of the heat extraction in
3D-EGS with thermal-hydraulic-mechanical coupling method based on discrete fractures model, Geothermics 74 (2018) 19–34,
http://dx.doi.org/10.1016/j.geothermics.2017.12.005.
[38] X. Song, Y. Shi, G. Li, R. Yang, G. Wang, R. Zheng, J. Li, Z. Lyu, Numerical simulation of heat extraction performance in enhanced
geothermal system with multilateral wells, Appl. Energy 218 (2018) 325–337, http://dx.doi.org/10.1016/j.apenergy.2018.02.172.
[39] S. Salimzadeh, A. Paluszny, H.M. Nick, R.W. Zimmerman, A three-dimensional coupled thermo-hydro-mechanical model for deformable
fractured geothermal systems, Geothermics 71 (2018) 212–224, http://dx.doi.org/10.1016/j.geothermics.2017.09.012.
[40] S.N. Pandey, A. Chaudhuri, S. Kelkar, A coupled thermo-hydro-mechanical modeling of fracture aperture alteration and reservoir
deformation during heat extraction from a geothermal reservoir, Geothermics 65 (2017) 17–31, http://dx.doi.org/10.1016/j.geothermics.
2016.08.006.
[41] A. Ghassemi, X. Zhou, A three-dimensional thermo-poroelastic model for fracture response to injection/extraction in enhanced
geothermal systems, Geothermics 40 (2011) 39–49, http://dx.doi.org/10.1016/j.geothermics.2010.12.001.
[42] Y. Wang, G. Ma, F. Ren, T. Li, A constrained Delaunay discretization method for adaptively meshing highly discontinuous geological
media, Comput. Geosci. 109 (2017) 134–148, http://dx.doi.org/10.1016/j.cageo.2017.07.010.
[43] G. Ma, Y. Wang, T. Li, Y. Chen, A mesh mapping method for simulating stress-dependent permeability of three-dimensional discrete
fracture networks in rocks, Comput. Geotech. 108 (2019) 95–106, http://dx.doi.org/10.1016/j.compgeo.2018.12.016.
[44] C. Yan, Y.-Y. Jiao, H. Zheng, A fully coupled three-dimensional hydro-mechanical finite discrete element approach with real porous
seepage for simulating 3D hydraulic fracturing, Comput. Geotech. 96 (2018) 73–89, http://dx.doi.org/10.1016/j.compgeo.2017.10.008.
[45] A. Lisjak, P. Kaifosh, L. He, B.S.A. Tatone, O.K. Mahabadi, G. Grasselli, A 2D fully-coupled, hydro-mechanical, FDEM formulation
for modelling fracturing processes in discontinuous, porous rock masses, Comput. Geotech. 81 (2017) 1–18, http://dx.doi.org/10.1016/
j.compgeo.2016.07.009.
[46] F. Ren, G. Ma, Y. Wang, T. Li, H. Zhu, Unified pipe network method for simulation of water flow in fractured porous rock, J. Hydrol.
547 (2017) 80–96, http://dx.doi.org/10.1016/j.jhydrol.2017.01.044.
[47] F. Ren, G. Ma, Y. Wang, L. Fan, Pipe network model for unconfined seepage analysis in fractured rock masses, Int. J. Rock Mech.
Min. Sci. 88 (2016) 183–196, http://dx.doi.org/10.1016/j.ijrmms.2016.07.023.
 Y. Wang, T. Li, Y. Chen et al. / Computer Methods in Applied Mechanics and Engineering 356 (2019) 465–489 489

[48] F. Ren, G. Ma, Y. Wang, L. Fan, H. Zhu, Two-phase flow pipe network method for simulation of CO 2 sequestration in fractured
saline aquifers, Int. J. Rock Mech. Min. Sci. 98 (2017) 39–53, http://dx.doi.org/10.1016/j.ijrmms.2017.07.010.
[49] Y. Chen, G. Ma, H. Wang, Heat extraction mechanism in a geothermal reservoir with rough-walled fracture networks, Int. J. Heat
Mass Transfer 126 (2018) 1083–1093, http://dx.doi.org/10.1016/j.ijheatmasstransfer.2018.05.103.
[50] Y. Chen, G. Ma, H. Wang, T. Li, Evaluation of geothermal development in fractured hot dry rock based on three dimensional unified
pipe-network method, Appl. Therm. Eng. 136 (2018) 219–228, http://dx.doi.org/10.1016/j.applthermaleng.2018.03.008.
[51] Y. Chen, G. Ma, T. Li, Y. Wang, F. Ren, Simulation of wormhole propagation in fractured carbonate rocks with unified pipe-network
method, Comput. Geotech. 98 (2018) 58–68, http://dx.doi.org/10.1016/j.compgeo.2017.11.009.
[52] G. Ma, Y. Chen, Y. Jin, H. Wang, Modelling temperature-influenced acidizing process in fractured carbonate rocks, Int. J. Rock Mech.
Min. Sci. 105 (2018) 73–84, http://dx.doi.org/10.1016/j.ijrmms.2018.03.019.
[53] Y. Chen, G. Ma, H. Wang, T. Li, Y. Wang, Application of carbon dioxide as working fluid in geothermal development considering a
complex fractured system, Energy Convers. Manag. 180 (2019) 1055–1067, http://dx.doi.org/10.1016/j.enconman.2018.11.046.
[54] Y. Chen, G. Ma, Y. Jin, H. Wang, Y. Wang, Productivity evaluation of unconventional reservoir development with three-dimensional
fracture networks, Fuel 244 (2019) 304–313, http://dx.doi.org/10.1016/j.fuel.2019.01.188.
[55] Y. Chen, G. Ma, H. Wang, The simulation of thermo-hydro-chemical coupled heat extraction process in fractured geothermal reservoir,
Appl. Therm. Eng. 143 (2018) 859–870.
[56] G. Ma, T. Li, Y. Wang, Y. Chen, Numerical simulations of nuclide migration in highly fractured rock masses by the unified pipe-network
method, Comput. Geotech. 111 (2019) 261–276, http://dx.doi.org/10.1016/j.compgeo.2019.03.024.
[57] J.C. Heinrich, P.S. Huyakorn, O.C. Zienkiewicz, A.R. Mitchell, An ‘upwind’ finite element scheme for two-dimensional convective
transport equation, Internat. J. Numer. Methods Engrg. 11 (1977) 131–143, http://dx.doi.org/10.1002/nme.1620110113.
[58] H.K. Versteeg, W. Malalasekera, An Introduction to Computational Fluid Dynamics: The Finite Volume Method, Pearson Education,
2007.
[59] W. Kimio, N. Yuichi, H. Toshiyuki, Numerical study on heat extraction from supercritical geothermal reservoir, in: World Geotherm.
Congr., Kyushu-Tohoku, Japan, 2000, pp. 3957–3961.
[60] F.C. Lai, F.A. Kulacki, The effect of variable viscosity on convective heat transfer along a vertical surface in a saturated porous
medium, Int. J. Heat Mass Transfer 33 (1990) 1028–1031, http://dx.doi.org/10.1016/0017-9310(90)90084-8.
[61] Z. Zhao, On the heat transfer coefficient between rock fracture walls and flowing fluid, Comput. Geotech. 59 (2014) 105–111,
http://dx.doi.org/10.1016/j.compgeo.2014.03.002.
[62] C. Xu, P. Dowd, Q. Li, Carbon sequestration potential of the habanero reservoir when carbon dioxide is used as the heat exchange
fluid, J. Rock Mech. Geotech. Eng. 8 (2016) 50–59, http://dx.doi.org/10.1016/j.jrmge.2015.05.003.
[63] A. McMahon, S. Baisch, Case study of the seismicity associated with the stimulation of the enhanced geothermal system at Habanero,
Australia, in: Proceedings, 2013, pp. 29–36.
[64] D. Chen, D. Wyborn, Habanero field tests in the Cooper Basin, Australia: a proof-of-concept for EGS, Geotherm. Resour. Counc.
Trans. 33 (2009) 159–164.
[65] E.M. Llanos, S.J. Zarrouk, R.A. Hogarth, Numerical model of the habanero geothermal reservoir, australia, Geothermics 53 (2015)
308–319, http://dx.doi.org/10.1016/j.geothermics.2014.07.008.
[66] C. Xu, P.A. Dowd, Z.F. Tian, A simplified coupled hydro-thermal model for enhanced geothermal systems, Appl. Energy 140 (2015)
135–145, http://dx.doi.org/10.1016/j.apenergy.2014.11.050.
[67] C. Xu, P.A. Dowd, R. Mohais, Connectivity analysis of the habanero enhanced geothermal system, in: Proc. 37th Work. Geotherm.
Reserv. Eng., 2012.