See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/267985873

Geothermal Fluid Dynamics

Conference Paper · December 2010

CITATIONS READS

2 1,818

1 author:

Michael O'Sullivan
University of Auckland
284 PUBLICATIONS   3,266 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

NCG reinjection View project

Transient wellbore simulation View project

All content following this page was uploaded by Michael O'Sullivan on 25 August 2015.

The user has requested enhancement of the downloaded file.

17th Australasian Fluid Mechanics Conference
Auckland, New Zealand
5-9 December 2010

Geothermal Fluid Dynamics

M.J. O’Sullivan

Department of Engineering Science

University of Auckland, Auckland 1142, New Zealand

Abstract implicit time differencing and upstream weighting of fluid

properties at block interfaces.
This article reviews the development of geothermal fluid
dynamics from the early studies of convection in a porous Over the last 30 years computer models of many geothermal
medium in the 1940s up to the present complex numerical fields have been set up and modelling has become an almost
models of geothermal fields. The early studies, and many others standard tool in the planning and management of geothermal
since, consider a simple problem with constant fluid properties projects. The author recently reviewed the state-of-the-art of
except in the buoyancy term where linear dependence of density geothermal modelling [62] and gave a historical perspective on
on temperature is assumed (the Boussinesq approximation). the development of geothermal modelling [63]. A brief
Even with this simple model complex behaviour is observed with discussion of the numerical techniques used in geothermal
a transition from conduction to convection at a Rayleigh number simulators, and their application to the study of reservoir physics
(Ra) of 42. At higher values of Ra transitions from steady to and modelling real geothermal systems, is given below.
oscillatory and then to irregular convection occur.
One of the most difficult tasks faced by a geothermal modeller is
In the 1970s numerical methods for studying general geothermal the problem of model calibration. This involves choosing model
flows, including boiling water and steam, were developed, and parameters such as permeability and porosity, and model
have been applied to models of many geothermal systems over boundary conditions such as the location and strength of deep
the last 40 years. Because of the highly nonlinear behaviour of inflows of heat and mass, so that the model results match the
boiling zones in hot geothermal systems geothermal simulators data. Solving this inverse problem is a topic of great current
such as TOUGH2 [79] use implicit time stepping and upstream interest and is discussed further in the penultimate section. In the
weighting of mobilities. final section other important present and future research topics in
geothermal reservoir modelling are discussed.
Geothermal fluid dynamics is still a challenging research area
with many topics requiring further work. Some of those Governing equations
discussed here are: automatic model calibration or inverse
Mass and heat flow in a geothermal system can be represented by
modelling, double porosity models, chemical reactions, fluid-
the following generic conservation equations:
rock interaction and engineered geothermal systems.
Introduction Am
 .Fm  qm (1)
The study of geothermal fluid dynamics began with the work by t
Horton and Rogers [40] and Lapwood [47] on the onset of
convection in a porous medium. They proved the classical result Ae
that for a closed square box, convection occurs for a Rayleigh
 .Fe  qe (2)
t
number (Ra) greater than 42. Many other studies on the onset of
convection in a porous medium for various configurations and Here Am and Ae are the amount of mass and heat per unit volume,
boundary conditions have been carried out over the last 65 years Fm and Fe are the mass flux and heat flux and qm and qe are the
since the work by Horton and Rogers. The book by Nield and amount of mass and heat produced or injected by wells.
Bejan [57] provides a good summary of this research.
For single-phase hot water Am and Ae are defined by
Post-onset behaviour was investigated by several researchers
using perturbation analysis [66],[59],[85], but in order to Am  l (3)
investigate flows for Ra well above the critical value of 42,
numerical methods had to be used. These included spectral Ae  (1   )  r crT  l ul (4)
methods [16],[42],[43],[67],[89],[91], finite difference methods
[8],[37],[38],[87] and finite element methods [17]. The results Here  is the porosity, l is the density of water, ul is the internal
are discussed further below. energy of water, r, cr are the rock density and rock specific heat
The research on the onset of convection and post-onset behaviour respectively and T is temperature. One of the important
all involved simple permeability structures and approximate fluid properties of geothermal systems is that most of the heat energy
properties and therefore had limited application to modelling the is contained in the rock matrix rather than the fluid and typically
flow of heat and mass in real geothermal fields which have very the first term in equation (4) accounts for 70 to 80% of the total
complex geological structures and a wide range of temperatures. energy content. In equations (3) and (4) and below the subscript l
is used to denote the properties of liquid water. Similarly the
Numerical methods suitable for modelling non-isothermal, multi- subscript v is used below to denote the steam or vapour phase.
phase, multi-component flow in heterogeneous, anisotropic
porous media developed during the 1970s. A number of For a two-phase mixture of boiling water and steam, equations
geothermal simulators were developed, most of which used fully (3) and (4) become
 Am   ( l Sl  v Sv ) (5) One of the interesting consequences of equations (11) and (12) is
the possibility of counter-flow with the flux of rising steam
matching the flux of water trickling downwards, giving zero nett
Ae  (1   )  r crT   ( l ul Sl   vuv Sv ) (6) mass transfer but a significant upward energy flow. Counter-flow
is the main mechanism for heat transfer in geothermal systems
Here Sl and Sv are the liquid and vapour saturations (volume with large boiling zones and a small through-flow of mass. This
fractions), v is the density of steam and uv is the internal energy is the case for vapour-dominated systems such as The Geysers in
of steam. California [100] or Kamojang [92] and Darajat [3],[35] in
For single phase flow of hot water the mass flux Fm is defined by Indonesia.
Darcy’s Law Onset of convection
k In the simple analysis described here the Bousinesq
Fm  (p  l g) (7) approximation is made, i.e. the density of water is assumed
l constant except in the buoyancy term where it is defined by:
Here k is permeability, l is the kinematic viscosity, p is pressure  l   0 [1   (T  T0 )] (13)
and g is the acceleration due to gravity.
Then the energy flux is given by the sum of an advective and Then, in the absence of external sinks or sources (i.e. wells),
conductive term: equations (1) and (2) can be simplified and written in terms of the
volume flux Q. For 2D flows with Q=(u,v,0) they become:
Fe  hl Fm  KT (8)
u v
 0 (14)
Here hl is the enthalpy of hot water and K is the effective thermal x y
conductivity of the saturated porous medium.
To represent two-phase flow, relative permeabilities krl and krv T T T   2T  2T 
are introduced and the mass flux of water and steam, Fml and Fmv,  u v  2  2  (15)
are calculated separately: t x y  x y 
kkrl Here  is the ratio of the thermal mass of the saturated rock
Fml  (p  l g) (9) matrix to that of water:
l
  [(1   )  r cr   0 c0 ]  0 c0 (16)
kkrv
Fmv  (p   v g ) (10) The specific heat of water c0 is assumed to be constant. The
v diffusivity  is given by:
The difficulty with using equations (9) and (10) is that the
relative permeability functions are not known and cannot be   K  0 c0 (17)
measured in the field. Most experimental studies of krl and krv
In deriving equations (14) and (15) the assumption is made that
have involved mixtures of liquid and gaseous hydrocarbons
the dependence of ul and hl on pressure is negligible.
rather than steam and water. It is known that at low saturations
water is trapped in the fractures and pores and becomes Using equation (13) Darcy’s law can be written as:
immobile, but the precise value of the residual or immobile water
saturation is not known and cannot be measured in situ. k p
Conversely at very high water saturations the steam bubbles u (18)
become trapped and the steam is immobile. One of the few  x
experimental studies that considered boiling flows [98] suggested
that straight-line relative permeabilities may be appropriate for k  p 
geothermal flows. v    0 g[1   (T  T0 )]  (19)
  y 
Various versions of the relative permeability functions have been
used in the geothermal context and are available in geothermal The classical onset problem considers a closed box with a
simulators such as TOUGH2 [78]. temperature T1 on the base and T0 on the top. In order to use
analytic methods to solve equations (14), (15), (18) and (19) for
One of the important issues in the mathematical modelling of
this problem it is necessary to assume that all parameters are
geothermal flows is how well equations (9) and (10), the two-
constant with respect to pressure and temperature. A few
phase version of Darcy’s Law, represent the flow of water and
previous studies [41],[90] have included the more physically
steam through a fractured rock matrix. Some modellers have
adopted a dual-porosity approach [4],[44], and in order to match realistic temperature dependence of the viscosity l but this
pressures, enthalpies and tracer test results dual-porosity, dual- modification requires the use of numerical methods.
permeability models have been used [68]. It is convenient for solving the onset problem to introduce the
The total mass and energy flows are given by following nondimensional variables:

Fm  Fml  Fmv (11) x*  x a , y*  y a , t*   t a 2

Fe  hl Fml  hv Fmv  KT (12) u*  au  , v*  av 

Here hl and hv are the enthalpies of water and steam respectively. p*   p  [ p0  0 g ( y  a )] k 
 T *  T  T0  T1  T0  The smallest (critical) value Racr=42 is given by a square single
cell with m=1, n=1 and l=1.
Here a is the thickness of the porous layer. A similar analysis was carried out for other sets of boundary
The introduction of a stream function also helps: conditions by Nield [56].

 *  * Post-onset behaviour
u*  , v*   For a Rayleigh number not much greater than the critical value
y * x * perturbation analysis can be applied [66],[85],[59]. A small
parameter defined by:
Now the governing equations can be reduced to two coupled
equations for the nondimensional temperature and stream
function, Tand(dropping the * for convenience):
 2  ( Ra  Racr ) Ra
Then expansions for  and T are assumed in the form:
 2  2 T
 2   Ra (20)
x 2
y x    (1)   2 ( 2)   3 (3)  ...

T  T  T  2T  2T T  1  y  T (1)   2T ( 2 )   3T ( 3)  ...
     (21)
t y x x y x 2 y 2 O’Sullivan and McKibbin [59] used expansions truncated at
O(6) and obtained results for Nusselt number vs cell-width that
The Rayleigh number Ra is defined by: agreed well with numerical results for Ra up to 200.

 (T1  T0 ) 0 gka The closed box convection problem discussed here is one of the
Ra  (22) simplest examples of hydrodynamic instability but even so it
l exhibits interesting and complex post-onset behaviour. An early
investigation by Straus [89] used stability analysis based on
The nondimensional boundary conditions for the closed box Fourier series representations of the unknowns, to prove two-
problem are shown in Fig. 1. dimensional convection is unstable for Ra>380. Caltigirone [8]
proved a similar result that fluctuating convection occurs for Ra
>384+/-5, and he also showed that the transition from steady
y convection to fluctuating convection is strongly dependent on
T  0,  0 cell-width (the length l in Fig. 1). Fig. 2 shows some of
1 Caltigirone’s results, for Nusselt number (Nu) vs cell-width for
Ra = 800. The plot shows that stable convection ceases for a
cell-width > 0.5.
T x  0, T x  0, Aidun and Steen [2] determined that for unicellular two-
 0  0 dimensional flow a Hopf bifurcation occurs at Ra =390.7 that
destabilizes the flow. This result was confirmed by Riley and
Winter [84] who also investigated the effect of cell-width. They
found that the flow becomes less stable as the cell becomes
x
0 T  1,  0 l wider, thus confirming the numerical results of Caltigirone [8].

10
Figure 1. Boundary conditions for the nondimensional closed box
problem
8
The onset problem is then easily solved by assuming steady flow
Nusselt number

and using linear stability analysis [57]. The conduction solution 6

is perturbed using

   (1) , T  1  y  T (1) (23)

4

Stable Nu
Equations (20) and (21) are easily solved using (23) giving 2 Max Nu
solutions: Min Nu

0
mx
 (1)  A sin sin ny
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Cell-width
l
mx
T (1)  B cos sin ny Figure 2. Nusselt number vs cell-width for Ra=800 (see Caltigirone [8]).
l Kimura et al. [42] used a pseudo-spectral numerical scheme to
With study two-dimensional, unicellular, time-dependent convection in
a square box. They found that with increasing Ra the flow
 2 m 2 l 2  n 2  evolves from steady S to chaotic (non-periodic) NP through a
2

Ra  sequence of bifurcations S  P(1)  QP2  P(2)  NP.

m2 l 2 Here P(1) and P(2) are periodic regimes and QP2 is a quasi-periodic
state with two basic frequencies. The transitions given by
Kimura et al. [42] are given in Table 1.
A study by Holzbecher [37] investigating the stability of higher researchers developing geothermal reservoir simulators
modes also confirmed the instability of the unicellular mode at [51],[70],[71],[72],[54]. The author and co-workers have worked
Ra = 390. Further it showed that transition to the oscillatory on a sequence of models that have increased in complexity as
regime occurs at Ra ~ 510 for the second mode and Ra ~ 970 for data have become available and as the capacity of the hardware
the third mode. has improved [6],[49],[50],[61],[63]. Our latest model [65] has
~10,000 blocks and another with ~30,000 blocks is under
Transition Ra development. This is still small compared with oil and gas
Conduction to Convection (S) 42 reservoir models which may contain ~106 blocks.

S to P(1) 380 - 400 Model Calibration

P(1) to QP2 500 - 520 The most difficult task in geothermal reservoir modelling is
model calibration. In mathematical terms model calibration is an
QP2 to P(2) 560 - 570 inverse problem which requires the choice of model parameters
(2) such as permeability and porosity so that a best fit is obtained
P to NP 850 - 1000 between the model results and measured data. Model calibration
involves three stages: conceptual modelling, natural state
Table 1. Rayleigh number for various transitions (see Kimura et al. [42]) modelling and production history matching [60].
Holzbecher found that for supercritical flows unicellular first Conceptual modelling is the term given to the process of
mode behaviour is only relevant for low values of Ra and then collecting all the data (geological, geophysical, geochemical etc)
second or third mode steady convection is preferred to oscillating and synthesizing it into two or three sketches of the hydro-
first mode convection. geological structure of the system.
The situation becomes even more complex if three-dimensional In natural state modelling the conceptual model is used to set up
flow is considered [36],[102],[39],[91],[87]. For example Straus the permeability structure of a computer model. Also the
and Schubert [91] investigated convection in a cubic box for Ra < locations and magnitudes of the deep sources of heat and mass at
150 and found that both 2D and 3D steady convection can occur the base of the computer model are determined from the
depending on the initial conditions, even though the Nusselt conceptual model.
number is greater for 2D flows for Ra < 97 but greater for 3D
flows for Ra > 97. Recently Sezai [87] investigated convection The computer model is run until stable steady-state conditions are
in a cube for Ra up to 1000. He identified ten steady flow achieved and then the model results are compared with data. In
patterns of which five show oscillatory behaviour in some particular, measured down-hole temperature profiles are
Rayleigh-number range. Further Sezai found that two of the compared with the model results and the locations of surface
steady solutions are stable for Ra up to 900, whereas an earlier outflows of heat and mass in the model are compared with the
study by Kimura et al. [43] had suggested that oscillations begin actual locations of hot springs and steaming ground. This aspect
at Ra=575. of model calibration is different for geothermal reservoirs than
for oil and gas reservoirs because of the dynamic, convective
Numerical Methods nature of geothermal systems. The convective system in a
Numerical methods are required for the solution of equations (1), geothermal system is affected by the permeability structure and
(2), (9) and (10), together with an equation of state (EOS) that thus influences the sub-surface temperature distribution.
includes accurate thermodynamic properties of water, and allows If the model temperature distribution does not produce a good
for the possibility of boiling i.e., the development of two-phase match to the measured down-hole temperatures, then the
conditions. permeability structure is adjusted and the model is run again.
Work began in the 1970s on numerical techniques for geothermal Many iterations of this process may be required. A plot of the
reservoir simulation [7],[11],[19],[20],[26],[48],[52],[54],[74], model results for one well at Wairakei are shown in Fig. 3.
[92]. A brief discussion of this development is given in the
review by the author [63]. A code comparison project organized
by the US Department of Energy in 1980 [88] showed that
several simulators produced similar results for a suite of
challenging test problems. The key numerical techniques
required to model the phase of transitions and strong advection
that occur in geothermal flows are fully implicit time differencing
and upstream weighting of interface quantities. The well-known
geothermal simulators such as TOUGH2 [78], FEHM [105],
STAR [73] and TETRAD [96] all use these techniques.
Upstream weighting provides a robust numerical technique but
unfortunately also adds numerical dispersion and the smearing of
sharp fronts [58]. Some work has been done on higher order
methods [75],[76] but they are difficult to generalize for Figure 3. Natural state temperatures in one well for the Wairakei model
unstructured three-dimensional finite volume grids. Similarly (data - blue symbols, model - red line)
good results have been achieved with an Euler-Lagrange
Once a reasonable natural-state model has been obtained the
approach for single-phase 2D flow [13] but more work is
results are used as the initial conditions for a simulation of the
required to implement the method for two-phase or 3D flows.
production and injection history, with the measured mass flows
Reservoir simulation has now been applied to setting up models being assigned to the appropriate model blocks. Then the
of many geothermal systems [60],[62]. One of the most fully pressure and enthalpy changes predicted by the model are
studied systems is Wairakei, New Zealand. Because of the ready compared to the data, and adjustments are made to the
availability of data Wairakei was used early on as a test case by permeabilities and porosities to improve the model. Typical
results for the Wairakei-Tauhara model [65] are shown in Figs. 4 The use of deeper models leads to the need for a thermodynamic
and 5. EOS that can handle higher pressures and temperatures.
Croucher and O’Sullivan [14] have implemented the IAPWS-97
Most models are calibrated by heuristic manual methods with the thermodynamic formulation [97], including a supercritical
experience and skill of the modeller determining the quality of capability valid for pressures up to 100 MPa and temperatures up
the model. to 800˚C. This improvement allows for models of high
temperatures and pressures to be set up, provided that the fluid
can be approximated as pure water. It would be very useful for
models of other fields (such as Ohaaki and Ngawha in New
Zealand) to have an EOS for mixtures of water, carbon dioxide,
and sodium chloride that is accurate for temperatures and
pressures ranging from atmospheric up to supercritical (pure
water) conditions. This would require the extension of the range
of validity of the ECO2N fluid property module [78] in
TOUGH2, which has been used for modelling carbon
sequestration [80],[81].

Figure 4. Pressure decline in the western borefield for the Wairakei Near surface behaviour
model (data - blue symbols, model - red line) The development of geothermal systems usually affects surface
features such as geysers, hot springs and steaming ground. For
example the exploitation of Wairakei caused the liquid-fed
features like geysers and hot springs to disappear but caused an
increase in the activity of steam heated features.
In order to model the interaction between the deep reservoir and
surface activity in the model of Wairakei, an air-water EOS is
used and the model is extended up to the ground surface [65].
Some other geothermal models take the water table as the top of
the model. The inclusion of the unsaturated zone in the model
works satisfactorily, but the movement of the water table is not
tracked very accurately as the minimum layer thickness is 25 m.
It would be useful to be able to handle the movement of the water
table in a geothermal model similarly to the way unconfined
Figure 5. Enthalpy changes for one well in the Wairakei model (data—
blue symbols, model—red line) aquifers are included in groundwater models. However a more
sophisticated approach is required as the surface, where the
pressure is atmospheric, may be partly boiling. Nevertheless,
Future directions in geothermal modelling
having the top surface of either water or steam able to move up or
This topic was discussed by the author at the TOUGH down through a grid-block would be a very useful advance for
Symposium in 2009 [64] and is briefly summarised again here. accurately representing near-surface behaviour, such as the
development of large areas of steaming ground.
Wellbore – reservoir interaction
Fluid-rock interaction
Many geothermal wells have more than one feed zone and the
proportion from each feed may vary as the reservoir pressures One of the great hopes of the geothermal industry is the
change. To represent these processes a combined well- development of enhanced (or engineered) geothermal systems
bore/reservoir simulator is required. A few researchers have (EGS) – also known as hot dry rock projects (HDR). The idea
experimented with this approach [5],[33], but it has not been very with an EGS is to drill one well, use hydraulic fracturing to create
successful as the maximum time-step achievable is controlled by a permeable zone and then drill a second well to intersect the new
the well-bore flow and is often very small. An alternative man-made permeability [103]. Then heat can be extracted from
approach using a multi-layer deliverability option is available in the fractured rock by circulating water down one well and up the
TOUGH2 but it assumes an approximate hydrostatic pressure other. It is a simple idea, but although there have been several
profile in the well. A method that is more accurate than this but attempts to establish pilot projects none of them have become
which is more computationally efficient than a full well- commercially successful. However there are very large resources
bore/reservoir model is required. of hot, low permeability, rock available worldwide [94] and EGS
are currently under development in many countries (e.g. [30]),
Larger-deeper models including Australia [9],[12],[31],[32],[83].
Very few models of geothermal systems include the whole of the The main difficulty with an EGS is in establishing a large enough
large-scale convective system. Thus, the base boundary fracture network to give a satisfactory heat exchange system
condition must include some input of very hot water, underground. In order to model the development of the fractures
corresponding to the upflow zone of the convective plume. It created by hydraulic fracturing and to model their performance as
would be better to make the model large enough so that the a heat exchanger, during the life of the EGS, it is necessary to be
whole convective system is contained in the model, in which case able to model the combined fluid flow, heat flow and rock
the permeability structure has to be compatible with the flow and mechanics problem [33].
temperature structure. Recently the model of Wairakei-Tauhara
developed by the author and co-workers has been extended by The author and others [77],[86] have considered the much easier
adding extra layers, so that it is now 4 km deep. Probably more but related problem of modelling subsidence in a geothermal
layers, extending the model down to 6–7 km, should be added system. By using temperature and pressure charges calculated
and a larger area included. with TOUGH2 as input for a rock mechanics simulation using
the ABAQUS package [1] we have had some success in
matching the occurrence of the subsidence bowls at Wairakei- is currently impractical. The simpler approach, and that which is
Tauhara [101]. The result for the Wairakei subsidence bowl is usually used with iTOUGH2, is to assign a relatively small
shown in Fig. 6. number of rock types and then use the permeabilities and
porosities of a subset of these rock-types as the parameters to be
The EGS problem is much more challenging than the subsidence optimized [45],[69]. However, even if the optimal values for all
problem and remains to be solved, although some progress was parameters, for all rock-types, are determined by iTOUGH2, the
made in a study of the Hijiori hot dry rock site [93], carried out resulting model is probably not going to be the best possible.
with the FEHM code [105].
It might be possible to produce a better model by subdividing the
zone assigned, say, to rock-type IGNIM into two new zones,
labelled IGNIA and IGNIB, for example. Then iTOUGH2 could
be re-run optimizing the parameters for IGNIA and IGNIB
independently. We have had some success with this technique,
but what is required is a more systematic approach with, say, an
outer XiTOUGH2 code that controls the re-assignment of rock-
types and calls iTOUGH2 to optimize parameter values for each
new rock-type structure.
There are some fundamental problems with inverse modeling
based on a least-squares-error approach (see Fox [27]). One
difficulty is that a global optimum or even a local optimum found
by a nonlinear optimization technique may not be a “good
solution,” in the sense that the optimal parameter values may not
be what a reservoir engineer expects or finds acceptable. What is
really required is to identify a region of the multidimensional
parameter space where good solutions are likely to be found.

Figure 6. Model results for the surface deformation at the Wairakei Recently Cui [15] has used statistical sampling based on the
subsidence bowl. MCMC method to calibrate a geothermal model. Statistical
sampling methods have the potential to produce better models as
Some interesting fluid-rock interaction studies of mass transfer they provide statistics of the parameters rather than just permit
through the ductile zone, below geothermal systems, have been estimates. As well as calibrating a 3D model Cui compared his
carried out by Fussels et al. [28] and Regenauer-Lieb et al. [82]. MCMC results with the results obtained using iTOUGH2 for a
Model calibration simple model consisting of a uniform layer feeding a geothermal
well. Measurements of pressure and production enthalpy for an
The major challenge facing the geothermal modelling community extended well test lasting 130 days were used to calibrate the
is automatic model calibration. The three approaches currently model. The parameters determined by calibration were
being investigated at the University of Auckland (and elsewhere) permeability, porosity, initial reservoir pressure, initial vapour
are: heuristic manual methods, inverse modelling and statistical saturation, relative permeability parameters.
sampling based on the multi-chain Monte Carlo method
(MCMC). A comparison of the MCMC and iTOUGH2 results is shown in
Fig. 7. It shows the probability distributions for porosity and
To speed up manual calibration and to make it less dependent on permeability (log scale) obtained using MCMC sampling and
the modeler, an “expert system” approach may be useful—and point estimates obtained with iTOUGH2.
this is one of our current research themes. The idea is to codify
the various strategies followed by a modeller and to apply them
in a systematic fashion. For example, in a natural state model, if
block I is too hot, then the following steps should be followed:
 Check flow directions for all connections between block I
and other blocks.
 For flows into block I, if the neighboring block J is hotter,
then decrease the permeability of block J. If block J is colder
than block I, then increase the permeability of block J.
 Repeat for all blocks sending fluid into block I. 0 0.2 0.4 -15 -14.9
2
Several rules of this kind are currently used by modellers, but Φ[-] log10(k) [m ]
need to be formalized. There are many challenges to overcome
Figure 7. Statistical distributions of porosity and permeability obtained
in order to make such an expert system work, and there are using MCMC sampling [15] compared with point estimates (red)
several unanswered questions. For example: will it converge to a obtained using iTOUGH2 [22].
good solution in a reasonable time?
The results show that the point estimates for porosity and
The inverse modelling approach using nonlinear optimization permeability obtained by Finsterle et al. [22] using iTOUGH2
methods has been applied using software such as iTOUGH2 probably correspond to a local minimum and they are statistically
[21],[23],[24],[25],[25] and PEST [18]. poor results.
The difficulty with using iTOUGH2 (or any inverse modeling Further advances with the MCMC technique are required to make
code) in calibrating a geothermal model is the choice of the it practically useful for calibrating geothermal models.
variable parameters. At one extreme, each block in the model Currently, we are investigating the use of a hierarchy of models
could be assigned different x, y, z permeabilities and porosities. ranging from a coarse grid to a fine grid, and we are investigating
This would result in a huge number of unknown parameters and
adaptive delayed acceptance algorithms [10],[15]. As with [5] Bhat, A., D. Swenson and S. Gosavi, Coupling the Hola
inverse modelling techniques (e.g. iTOUGH2), MCMC is ideal wellbore simulator with TOUGH2, Proc. 30th Workshop on
for implementation on a cluster of computers in a distributed Geothermal Reservoir Engineering, Stanford University,
memory configuration. In the future, it may be possible to use a Stanford, Calif., Jan. 31 – Feb. 2, 2005.
cluster of multi-core processors, each running a parallelized
version of TOUGH2 [55],[103], to calibrate large complex [6] Blakeley, M.R. and O'Sullivan, M.J., Modelling production
geothermal models. and recharge at Wairakei. In: Proceedings of the 4th New
Zealand Geothermal Workshop, University of Auckland,
Conclusions 1982, 23-31.
The basic theoretical problems in geothermal fluid dynamics [7] Brownell, D.H., Garg, S.K., Pritchett, J.W., Governing
have received much attention during the last 70 years with many equations for geothermal reservoirs. Water Resources
studies carried out on the onset of convection and on the large Research, 13, 1977, 929-934.
amplitude post-onset behaviour. The problem of convection in a
porous medium heated from below is one of simplest examples [8] Caltigirone, J.P. Thermoconvective instabilities in a
of hydrodynamic instability but nevertheless it exhibits horizontal porous layer. J. Fluid Mechanics, 72, 1975, 269-
interesting complex behaviour, with several modes of convection 287.
possible each of which changes from steady to oscillatory flow at [9] Chen, D. Concepts of a basic EGS model for the Cooper
different values of Ra. Further research is required to fully Basin, Australia. In: Proc. World Geothermal Congress
describe all the transitions. 2010, Bali, Indonesia, 25-29 April 2010.
Since their first development in the 1970s techniques for [10] Christen, J.A. and Fox, C. MCMC using and approximation.
modelling flows in geothermal systems have developed J. Computational and Graphical Statistics, 14(4), 2005, 795-
considerably but more remains to be done. We would like to be 810.
able to run bigger and better models of geothermal fields; we
would like to be able to calibrate them better and more quickly; [11] Coats, K.H., Geothermal reservoir modeling. SPE paper
and we would like the models to be able to incorporate more 6892 presented at the 52nd Annual Fall Technical
complex reservoir behaviour. Conference and Exhibition, Society of Petroleum Engineers,
Denver, Colorado, USA, 1977.
Improved calibration methods. All three of the calibration
methods discussed above (heuristic manual methods, inverse [12] Cooper, G.T., Beardsmore, G.R., Waining, B.S., Pollington,
modelling, statistical sampling) require further research to make N. and Driscoll, J.P. The relative costs of engineered
them work well. geothermal system exploration and development in
Australia. In: Proc. World Geothermal Congress 2010, Bali,
Improved numerics. Our highest priority item in this category is Indonesia, 25-29 April 2010.
the tracking of the movement of the water/steam-table. This is
probably achievable, and Euler-Lagrange methods [13] could be [13] Croucher, A.E., O’Sullivan, M.J., Kikuchi, T. and Yasuda,
implemented in TOUGH2. We are less optimistic about Y., Eulerian-Lagrangian tracer simulation with TOUGH2.
implementing front tracking into general 3D unstructured grids Geothermics 33(4), 2004, 503-520.
commonly used in geothermal models. [14] Croucher, A.E. and M.J. O’Sullivan, Application of the
New EOSs. There are no fundamental impediments to the computer code TOUGH2 to the simulation of supercritical
development of EOS modules that can handle mixtures of water conditions in geothermal systems, Geothermics, 37, 2008,
and CO2 (and perhaps NaCl) over a wide temperature and 622-634.
pressure range, although working out the details could be time- [15] Cui, T. Bayesian calibration of geothermal reservoir models
consuming (e.g., Kissling et al. [46]). via Markov chain Monte Carlo. PhD thesis, University of
Fluid-rock interaction. The problem of predicting the spread of Auckland, New Zealand, 2010.
a fracture zone in an HDR (EGS) project is challenging. More [16] De La Torre Juarez, M. and Busse, F.H. Stability of two-
research is required. dimensional convection in a fluid-saturated porous medium.
Some of these aims can be met with current techniques, but J. Fluid Mechanics 292, 1995, 305-323.
others require more research. [17] Diersch, H.-J.G. and Kolditz, O. Variable-density flow and
References transport in porous media: approaches and challenges.
Advances in Water Resources, 25, 2002, 899-944.
[1] ABAQUS Inc., Pawtucket, RI, ABAQUS Analysis User’s
Manual Version 6.4, 2003. [18] Doherty, J., PEST: Model independent parameter
estimation, Watermark Numerical Computing, Corinda,
[2] Aidun, C.K. and Steen, P.H. Transition to oscillatory Australia, 2005. http://www.pesthomepagecom/
convection in a fluid-saturated porous medium. J.
Thermophys. Heat Transfer, 1, 1987, 268-273. [19] Faust, C.R., Mercer, J.W., Geothermal reservoir simulation,
1: Mathematical models for liquid- and vapor-dominated
[3] Alamsyah, O., Bratakusuma, B., Hoang, V. and Roberts, hydrothermal systems. Water Resources Research 15, 1979,
J.W. Dynamic modeling of Darajat Field using numerical 23-30.
simulation. In: Proceedings of the World Geothermal
Congress 2005, Antalya, Turkey, 24-29 April, 2005. [20] Faust, C.R., Mercer, J.W., Geothermal reservoir simulation,
2: Numerical simulation techniques for liquid- and vapor-
[4] Antunez, E.U., Bodvarsson, G.S. and Walters, M.A. dominated hydrothermal systems. Water Resources
Numerical simulation study of the northwest Geysers Research 15, 1979, 31-46.
geothermal field, a case study of the Coldwater Creek
steamfield. Geothermics, 23(2), 1994, 127-141. [21] Finsterle, S. and Pruess K. Solving the estimation-
identification problem in two-phase flow modelling. Water
Resources Research, 31(4), 1995, 913-924.
[22] Finsterle, S., Pruess, K., Bullivant, D.P. and O'Sullivan, [37] Holzbecher, E. On the relevance of oscillatory convection
M.J., Application of inverse modeling to geothermal regimes in porous media – review and numerical
reservoir simulation. In Proc. 22nd Workshop on experiments. Computers & Fluids, 30, 2001, 189-209.
Geothermal Reservoir Engineering, Stanford University,
Stanford, California, 1997, 309-316. [38] Horne, R.N. and O’Sullivan, M.J. Oscillatory convection in
a porous medium heated from below. J. Fluid Mechanics,
[23] Finsterle, S. Multiphase inverse modelling: review and 66, 1974, 339-352.
TOUGH2 applications. Vadose Zone J., 3, 198, 747-762.
[39] Horne, R.N. Three-dimensional natural convection in a
[24] Finsterle, S., iTOUGH2 User’s Guide, Report LBNL-40040, porous medium heated from below. J. Fluid Mechanics, 92,
Updated Reprint, Lawrence Berkeley National Laboratory, 1979, 751-766.
Berkeley, Calif., 2007.
[40] Horton, C.W. and Rogers, F.T. Convection currents in a
[25] Finsterle, S., iTOUGH2 Command Reference, Report porous medium. J. Appl. Phys., 16, 1945, 367-370.
LBNL-40041, Updated Reprint, Lawrence Berkeley
National Laboratory, Berkeley, Calif., 2007. [41] Kassoy, D.R. and Zebib, A. Variable viscosity effects on the
onset of convection in porous media. Physics of Fluids,
[26] Finsterle, S., iTOUGH2 Sample Problems, Report LBNL- 18(12), 1975, 1649-1651.
40042, Updated Reprint, Lawrence Berkeley National
Laboratory, Berkeley, Calif., 2007. [42] Kimura, S., Schubert, G. and Straus, J.M. Route to chaos in
porous-medium thermal convection. J. Fluid Mechanics,
[27] Fox, C.N. I’m sorry, but you have to do better than least- 166, 1986, 305-324.
squares if you want to make quantitative estimates.
Presentation at NZIMA Workshop on Energy, Wind and [43] Kimura, S., Schubert, G. and Straus, J.M. Time-dependent
Water: algorithms for simulation, optimization and control, convection in a fluid saturated porous cube heated from
University of Auckland New Zealand, February, 2009. below. J. Fluid Mechanics 207, 1989, 153-189.
http://www2.esc.auckland.ac.nz/EnergyWindWater/ [44] Kirukhin, A.V. and Yampolsky, V.A. Modeling study of the
[28] Fussels, F., K. Regenauer-Lieb, J. Lui, R.M. Hough and F. Pauzhetsky geothermal field, Kamchatka, Russia.
De Carlo, Creep cavitation can establish a dynamic granular Geothermics, 33, 2004, 421-442.
fluid pump in ductile shear zones, Nature, Letters, 459, [45] Kirukhin, A.V., Asaulova, N.. and Finsterle, S. Inverse
2009, 974-977. modeling and forecasting for the exploitation of the
[29] Garg, S.K. and Pritchett, J.W., On pressure-work, viscous Pazhetsky geothermal field, Kamchatka, Russia.
dissipation and the energy balance relation for geothermal Geothermics, 37, 2008, 540-562.
reservoirs. Advances in Water Resources Research 1, 1977, [46] Kissling, W.M., Transport of three-phase hyper-saline brines
41-47. in porous media: theory and code implementation, Transport
[30] Gerard, A., Genter, A., Kohl, T., Lutz, P., Rose. P. And in Porous Media, 61, 2005, 25-44.
Rummel, F. The deep EGS (Enhanced Geothermal System) [47] Lapwood, E.R. Convection of a fluid in a porous medium.
project at Soultz-sous-Forets (Alsace, France). Geothermics, Proc. Camb. Phil. Soc., 44, 1948, 508-521.
35, 2006, 473-483.
[48] Lasseter, T.J., Witherspoon, P.A., Lippmann, M.J., The
[31] Goldstein, B.A., Hill, A.J., Budd, A.R., Holgate, F. and numerical simulation of heat and mass transfer in
Malavazos, M. Hot rocks in Australia- national outlook. In multidimensional two-phase reservoirs. In: Proceedings of
Proc. 33rd Workshop on Geothermal Reservoir the Second United Nations Symposium on the Development
Engineering, Stanford University, Stanford, California, and Use of Geothermal Resources, San Francisco,
2008. California, USA, 3, 1975, 1715-1725.
[32] Goldstein, B.A., Hill, A.J., Long, A., Budd, A.R., Holgate, [49] Mannington, W.I., O’Sullivan, M.J. and Bullivant, D.P. An
F. and Malavazos, M. Hot rock geothermal energy plays in air/water model of the Wairakei-Tauhara geothermal system.
Australia. In Proc. 34th Workshop on Geothermal Reservoir In: Proceedings of the World Geothermal Congress 2000,
Engineering, Stanford University, Stanford, California, Kyushu-Tohoku, Japan, 28 May-10 Jun. 2000, 2713-2718.
2009.
[50] Mannington, W.I., O’Sullivan, M.J., Bullivant, D.P., and
[33] Hadgu, T., R.W. Zimmerman and G.S. Bodvarsson, Coupled Clotworthy, A.W. Reinjection at Wairakei-Tauhara: a
reservoir-wellbore simulations of geothermal reservoir modelling case study. In: Proceeding of the 29th Workshop
behavior, Geothermics, 24(2), 1995, 145-166. on Geothermal Reservoir Engineering, Stanford University,
[34] Hayashi, K., Willis-Richards, J., Hopkirk, R.J. and Niibori, January, 2004, 166-173.
Y. Numerical models of HDR geothermal reservoirs - a [51] Mercer, J.W., Pinder, G.F., Donaldson, I.G., A Galerkin
review of current thinking and progress. Geothermics, 28, finite element analysis of the hydrothermal system at
1999, 507-518. Wairakei, New Zealand. J. Geophys. Res., 80(17), 1975,
[35] Hoang, V., Alamsyah, O. and Roberts, J.W., 2005. Darajat 2608-2621.
geothermal field expansion performance – a probabilistic [52] Mercer, J.W. and Faust, C.R., Simulation of water and
approach. In: Proceedings of the World Geothermal vapor-dominated hydrothermal systems. Paper SPE5520,
Congress 2005, Antalya, Turkey, 24-29 April, 2005. presented at the 50th Annual Fall Meeting of the Society of
[36] Holst, P.H. and Aziz, K. Transient three-dimensional natural Petroleum Engineers, AIME, Dallas, Texas, USA, 1975.
convection in confined porous media. J. Fluid Mechanics, [53] Mercer, J.W., Faust, C.R., A review of numerical simulation
15, 1972, 73-90. of hydrothermal systems. Hydrological Sciences - Bulletin-
des Sciences Hydrologiques, 24(3), 1979, 335-343.
[54] Mercer, J.W. and Faust, C.R., Geothermal reservoir environmental effects of geothermal power production
simulation, 3: Application of liquid- and vapor-dominated phase IIA. Final Report No. SSS-R-77-2998, Systems,
hydrothermal modeling techniques to Wairakei, New Science and Software, California, USA, 1976.
Zealand. Water Resources Research, 15, 1979, 653-671.
[71] Pritchett, J.W., Rice, L.F., Garg, S.K., Reservoir
[55] Moridis, G.J., Kowalsky, M.B. and Pruess, K. engineering data: Wairakei geothermal field, New Zealand.
TOUGH+Hydrate v1.0, User’s Manual: A code for the Final Report No. SSS-R-78-3597. Systems, Science and
simulation of system behavior in hydrate bearing geologic Software, California, USA, 1978.
media, Paper LBNL-149E, Lawrence Berkeley National
Laboratory, Berkeley, California, 2008. [72] Pritchett, J.W., Rice, L.F. and Garg, S.K., Reservoir
simulation studies: Wairakei geothermal field, New
[56] Nield, D.A. Onset of thermohaline convection in a porous Zealand. Final Report, Geothermal Reservoir Engineering
medium. Water Resources Research, 4(3), 1968, 553-560. Management Program report GREMP-11, Lawrence
Berkeley Laboratory, Berkeley, California, USA, 1980.
[57] Nield, D.A. and Bejan, A. Convection in porous media.
Springer-Verlag, 1992. [73] Pritchett, J.W. STAR: A geothermal reservoir simulation
system. In: Proceedings World Geothermal Congress 1995,
[58] O'Sullivan, M.J. and Pruess, K., Analysis of injection testing Florence, Italy, 1995, 2959–2963
of geothermal reservoirs, Transactions of Geothermal
Resources Council Annual Meeting, 4, 1980, 401-404. [74] Pruess, K. and Schroeder, R.C., Geothermal reservoir
(Also published as Lawrence Berkeley Laboratory Report simulation with SHAFT79, In: Proceedings of the 5th
L.B.L. 10985). Workshop on Geothermal Reservoir Engineering, Stanford,
California, USA, 1979, 183-188.
[59] O'Sullivan, M.J. and McKibbin, R., Variation of heat
transfer with cell-width in a porous layer. In: Mathematics [75] Pruess, K. Grid orientation and capillary pressure effects in
and Models in Engineering Science, DSIR, Wellington, New the simulation of water injection into depleted vapor zones.
Zealand, 129-138, 1982. Geothermics 20(5/6), 1991, 257-277.
[60] O'Sullivan, M.J., Geothermal reservoir simulation. J. Energy [76] Pruess, K. Numerical simulation of water injection into
Research, 9, 1985, 319-332. vapor-dominated reservoirs. In: Proc. World Geothermal
Congress 1995, Florence, Italy, 1995, 1673-1678.
[61] O'Sullivan, M.J., Bullivant, D.P., Follows, S.E. and
Mannington, W.I. Modelling of the Wairakei–Tauhara [77] Pruess, K., Garcia, J., Kovscek, T,, Oldenburg, C., Rutqvist,
geothermal system. In: Proceedings of the TOUGH J., Steefel, C. and Xu, T. Code intercomparison builds
Workshop '98, Berkeley, California, USA, 1998, 1-6. confidence in numerical simulation models for geologic
disposal of CO2, Energy, 29, 2004, 1431-1444.
[62] O’Sullivan, M.J., Pruess, K. and Lippmann, M.J. State of the
art of geothermal reservoir simulation. Geothermics, 30, [78] Pruess, K. ECO2N: A TOUGH2 Fluid property module for
2001, 395-429. mixtures of water, NaCl and CO2, Report LBNL-57952,
Lawrence Berkeley National Laboratory, Berkeley, Calif.,
[63] O'Sullivan, M.J. A history of numerical modelling of the 2005.
Wairakei geothermal field. Geothermics, 38(1), 2009, 155-
168. [79] Pruess, K., C. Oldenburg, and G.J. Moridis, TOUGH2
User’s Guide, Version 2.0, Report LBNL-43134, Lawrence
[64] O’Sullivan, M.J. Future directions in geothermal modelling. Berkeley National Laboratory, Berkeley, California, 1999.
In: Proceedings TOUGH Symposium 2009, Lawrence
Berkeley National Laboratory, Berkeley, California, [80] Pruess, K. Leakage of CO2 from geologic storage: role of
September 14-16, 2009, 365-372. secondary accumulation at shallow depth, Int. J.
Greenhouse Gas Control, 2(1), 2008, 37–46.
[65] O’Sullivan, M.J. and Yeh, A. Wairakei-Tauhara Modeling
Report, Uniservices and Department of Engineering [81] Pruess, K. On CO2 fluid flow and heat transfer behavior in
Science, University of Auckland, 283 pp, 2010. (included as the subsurface, following leakage from a geologic storage
part of Contact Energy’s application for resource consents reservoir, Environmental Geology, 54(8), 2008, 1677–1686.
http://www.contactenergy.co.nz/web/pdf/environmental/P4
ReservoirModellingReport.pdf). [82] Regenauer-Lieb, K., B. Hobbs, A. Ord, O. Gaede and R.
Vernon, Deformation with coupled chemical diffusion,
[66] Palm, E., Weber, J.E. and Kvernvold, O. On steady Physics of the Earth and Planetary Interiors, 172, 2009, 43-
convection in a porous medium. J. Fluid Mechanics, 54 54.
1972, 153-161.
[83] Reid, P.W., Bendall, B.R. and McAllister, L. Developing
[67] Palm, E. Nonlinear thermal convection. Annual Review of large scale, base load EGS power – the Paralana project,
Fluid Mechanics, 7, 1975, 39-61. South Australia. In: Proc. World Geothermal Congress
2010, Bali, Indonesia, 25-29 April 2010.
[68] Pham, M., Klein, C., Ponte, C., Caberas, R., Martins, R. And
Rangel, G. Production/injection optimisation using [84] Riley, D.S. and Winters, K.H. Time-periodic convection in
numerical modeling at Ribeira Grande, Sao Miguel, Azores, porous media: the evolution of Hopf bifurcations with
Portugal. In: Proc. World Geothermal Congress 2010, Bali, aspect ratio, J. Fluid Mechanics, 223, 1991, 457-474.
Indonesia, 25-29 April 2010.
[85] Rubin, H. On the analysis of cellular convection in porous
[69] Porras, E.A., Tanaka, T., Fujii, H. and Itoi, R. Numerical media. Int. J. Heat Mass Transfer, 18, 1975, 1483-1486.
modeling of the Momotombo geothermal system,
Nicaragua. Geothermics, 36, 2007, 304-329. [86] Rutqvist, J. and C-F Tsang, TOUGH-FLAC: A numerical
simulator for analysis of coupled thermo-hydrologic-
[70] Pritchett, J.W., Garg, S.K., Brownell, D.Y., Rice, M.H., mechanical processes in fractured and porous geological
Riney, T.D., Hendrickson, R.R., Geohydrological media under multi-phase flow conditions, Proc. TOUGH
 Symposium 2003, Lawrence Berkeley National Laboratory, [96] Vinsome, P.K.W. and Shook, G.M. Multi-purpose
Berkeley, Calif., 2003. simulation. J. Petrol. Science Eng., 9(1), 1993, 29-38.
[87] Sezai, I. Flow patterns in a fluid-saturated porous cube [97] Wagner, W., J.R. Cooper, A. Dittman, J. Kijima, H.-J.
heated from below. J. Fluid Mechanics, 523, 2005, 393-410. Kretzschmar, A. Kruse, R. Mareš, K. Oguchi, H. Sato, I.
Stöcker, O. Šifner, Y. Takaishi, I. Tanishita, J. Trübenbach,
[88] Stanford Geothermal Program (ed.) Proceedings of the J. and Th. Willkommen, The IAPWS Industrial Formulation
special panel on geothermal model intercomparison study. 1997 for the Thermodynamic Properties of Water and
Report: SGP-TR42. Also summarised in Proc. 6th Steam, ASME J. Eng. Gas Turbines and Power, 122, 2000,
Workshop on Geothermal Reservoir Engineering, Stanford 150-182.
University, Stanford California, 1980, 253-257.
[98] Wang, C.T. and Horne, R.N. Boiling flow is a horizontal
[89] Straus, J.M. Large amplitude convection in porous media. J. fracture. Geothermics, 29, 2000, 759-772.
Fluid Mechanics, 64, 1974, 51-63.
[99] White, S.P., Creighton, A.L, Bixley, P.F. and Kissling,
[90] Straus, J.M. and Schubert, G. Thermal convection of water W.M. Modeling the dewatering and depressurization of the
in a porous medium: effects of temperature- and pressure- Lihir open-pit gold mine, Papua New Guinea. Geothermics,
dependent thermodynamic transport properties. J. Geophys. 33, 2004, 443-456.
Res. 82, 1977, 325-333.
[100] Williamson, K.H. Reservoir simulation of The Geysers
[91] Straus, J.M. and Schubert, G. Three-dimensional convection geothermal field. In Proc. 15th Workshop on Geothermal
in a cubic box of fluid saturated porous material. J. Fluid Reservoir Engineering, Stanford University, Stanford,
Mechanics, 91, 1979, 155-165. California, 1990, 113-123.
[92] Suryadarma, Azimuddin, T., Dwikorianto, T., and Fauzi, A. [101] Yeh, A. and O’Sullivan, M.J. Computer modelling of
The Kamojang geothermal field: 25 years operation. In: subsidence in geothermal fields. In: Proc. 29th NZ
Proc. World Geothermal Congress 2005, Antalya, Turkey, Geothermal Workshop, University of Auckland, 2007.
2005.
[102] Zebib, A. and Kassoy, D.R. Three-dimensional natural
[93] Tenma, N., Yamaguchi, T. and Zyvoloski, G.A. The Hijiori convection motion in a confined porous medium. Phys.
hot dry rock test site, Japan: evaluation and optimization of Fluids, 21, 1978, 1-3.
heat extraction from a two-layered reservoir. Geothermics,
37, 2008, 19-52. [103] Zhang, K., Moridis, G.J. and Pruess, K. TOUGH+CO2:
A multiphase flow simulator for CO2 geologic sequestration
[94] Tester, J.W., Anderson, B.J., Batchelor, A.S., Blackwell, in saline aquifers, In: Proceedings TOUGH Symposium
D.D., DiPippo, R., Drake, E.M., Garnish, J., Livesay, B., 2009, Lawrence Berkeley National Laboratory, Berkeley,
Moore, M.C., Nichols, K.M., Petty, S., Toksov, M.N., California, September 14-16, 2009, 41-48.
Veatch Jr, R.W., Baria, R., Augustine, C., Murphy, E.,
Negraru, P. and Richards, M. The future of geothermal [104] Zimmerman, G., Moeck, I. and Blocher, G. Cyclic
energy: impact of Enhanced Geothermal Systems (EGS) on waterfrac stimulation to develop an Enhanced Geothermal
the United States in the 21st century. Massachusetts Institute System (EGS) - Conceptual design and experimental results.
of Technology report, Cambridge, MA, USA, 2006, 372 pp. Geothermics, 39, 2010, 659-69.
[95] Toronyi, R.M., Farouq Ali, S.M., Two-phase, two- [105] Zyvoloski, G.A., Dash, Z. and Kelkar, S. FEHMN 1.0:
dimensional simulation of a geothermal reservoir. Society of Finite Element Heat and Mass Transfer Code, Rev. 1, Los
Petroleum Engineers Journal 17(3), 1977, 171-183. Alamos National Laboratory, Report LA-12062-MS, Los
Alamos, NM, USA, 1992.

View publication stats