Content uploaded by Niels Hartog
Author content
All content in this area was uploaded by Niels Hartog on Feb 05, 2018
Content may be subject to copyright.
PAPER
The use of salinity contrast for density difference compensation
to improve the thermal recovery efficiency in high-temperature
aquifer thermal energy storage systems
Jan H. van Lopik
1,2
&Niels Hartog
1,2
&Willem Jan Zaadnoordijk
2,3
Received: 5 September 2015 /Accepted: 6 January 2016
#Springer-Verlag Berlin Heidelberg 2016
Abstract The efficiency of heat recovery in high-temperature
(>60 °C) aquifer thermal energy storage (HT-ATES) systems is
limited due to the buoyancy of the injected hot water. This study
investigates the potential to improve the efficiency through com-
pensation of the density difference by increased salinity of the
injected hot water for a single injection-recovery well scheme.
The proposed method was tested through numerical modeling
with SEAWATv4, considering seasonal HT-ATES with four
consecutive injection-storage-recovery cycles. Recovery effi-
ciencies for the consecutive cycles were investigated for six
cases with three simulated scenarios: (a) regular HT-ATES, (b)
HT-ATES with density difference compensation using saline
water, and (c) theoretical regular HT-ATES without free thermal
convection. For the reference case, in which 80 °C water was
injected into a high-permeability aquifer, regular HT-ATES had
an efficiency of 0.40 after four consecutive recovery cycles. The
density difference compensationmethodresultedinanefficien-
cy of 0.69, approximating the theoretical case (0.76). Sensitivity
analysis showed that the net efficiency increase by using the
density difference compensation method instead of regular
HT-ATES is greater for higher aquifer hydraulic conductivity,
larger temperature difference between injection water and
ambient groundwater, smaller injection volume, and larger aqui-
fer thickness. This means that density difference compensation
allows the application of HT-ATES in thicker, more permeable
aquifers and with larger temperatures than would be considered
for regular HT-ATES systems.
Keywords Aquifer thermal energy storage (ATES) .Density
difference compensation .Groundwater density .Recovery
efficiency .Numerical modeling
Introduction
The rising demand for sustainable energy sources and CO
2
emis-
sion reduction has led to intensified use of seasonal aquifer ther-
mal energy storage (ATES) systems (Sanner et al. 2003). Thus
far, ATES is mainly used for seasonal heating and cooling of
buildings, where hot water is injected in the subsurface during
summers and extracted during winters, and vice versa. Currently,
the majority of ATES systems have limited temperature differ-
ences (ΔT< 15 °C) between the ambient groundwater on the one
hand, and the injected warm and cold water on the other hand.
The number of high temperature aquifer thermal energy storage
(HT-ATES) systems is still limited, although the storage of water
with higher temperatures (e.g. >70 °C) increases both the energy
storage capacity and overall energy efficiency (e.g. Kabus and
Seibt 2000;Sanneretal.2005; Réveillère et al. 2013). A huge
amount of waste heat is produced worldwide in a wide range of
industrial processes such as from incinerator and electricity
plants (Meyer and Todd 1973). Therefore, HT-ATES systems
can play a critical role in buffering the temporal mismatch
between (continuous) heat supply and (seasonal) demand.
To date, most of the technical challenges induced by the
high temperatures that hampered the early growth in the num-
ber of HT-ATES systems have now been resolved, such as
*Jan H. van Lopik
j.h.vanlopik@uu.nl
1
Department of Earth Sciences. Environmental Hydrogeology group,
Utrecht University, Budapestlaan 4, 3584
CD Utrecht, The Netherlands
2
KWR Watercycle Research Institute, Groningenhaven 7, 3433
PE Nieuwegein, The Netherlands
3
Water Resources Section, Faculty of Civil Engineering and
Geosciences, Delft University of Technology, Stevinweg 1, 2628
CN Delft, The Netherlands
Hydrogeol J
DOI 10.1007/s10040-016-1366-2
appropriate water treatment and material selection to prevent
the occurrence of mineral scaling and corrosion (Sanner et al.
2003). However, the economic feasibility of HT-ATES sys-
tems is largely determined by the thermal recovery efficiency,
and relative low thermal recoveries up to values of only 0.42
were obtained during HT-ATES field experiments conducted
in both confined and unconfined sandy aquifers (e.g. Mathey
1977;Palmeretal.1992;Molzetal.1979;1983a,b).
Besides heat loss due to thermal conduction to colder sur-
rounding formations (e.g. Doughty et al. 1982), free thermal
convection during HT-ATES negatively impacts thermal ener-
gy recovery (e.g. Buscheck et al. 1983;Molzetal.1983a). The
temperature difference between hot injection water and cold
ambient groundwater results in a net buoyancy difference.
Therefore, buoyancy forces cause upward flow of hot injection
water which results in tilting of the initially vertical hot water
front (Fig. 1a,b). Also, the viscosity of water at elevated tem-
peratures is lowered, resulting in enhanced free thermal convec-
tion (Hellström et al. 1979). Field experiments (e.g. Molz et al.
1983a) show that HT-ATES in high-permeability aquifers at
high temperature contrasts between injected water (81 °C)
and ambient groundwater (20 °C) result in significant free ther-
mal convection and hence a low thermal recovery (0.45).
The use of lower storage temperatures and the selection of
low-permeability target aquifers are currently seen as the main
design options for HT-ATES systems to reduce heat losses due
to density-driven flow (e.g. Doughty et al. 1982;Schoutetal.
2014). However, besides limiting the range of suitable aquifers
and the use of lower, energetically less attractive, storage tem-
peratures, the selection of low-permeability aquifers negatively
impacts the hydraulic capacity of a HT-ATES system, and there-
fore the heat storage capacity. Moreover, the selection of low-
permeability aquifers increases the risk of well clogging by par-
ticles (e.g. Olsthoorn 1982). Therefore, the work reported here
studies the possibility of minimizing free thermal convection in
HT-ATES systems by using saline water for heat storage to
compensate for the density difference with the ambient, cooler
and less saline groundwater. Potentially, this would significantly
increase recovery efficiencies and enable the use of higher per-
meability aquifers and higher injection temperatures for HT-
ATES systems. This study explored the potential of this ap-
proach with numerical density-dependent flow simulations of
(a) regular HT-ATES systems, (b) HT-ATES systems with den-
sity difference compensation using saline water, and (c)
theoretical regular HT-ATES cases that consider no free thermal
convection by neglecting density differences due to the temper-
ature contrast between hot injection water and cold ambient
groundwater. The latter scenario provided an upper bound for
the improvement obtainable with density difference compensa-
tion. Firstly, a sensitivity analysis was performed by simulating
multiple seasonal HT-ATES recovery cycles at various aquifer
and environmental conditions. Secondly, the density difference
compensation method was simulated for an actual HT-ATES
system, a pilot in a highly permeable aquifer (Molz et al.
1983a; Buscheck et al. 1983).
Methods
SEAWAT
To simulate water, heat and solute transport during the HT-
ATES recovery cycles, SEAWATv4 was used (Langevin et al.
2008; Guo and Langevin 2002). This code is a coupled version
of the groundwater flow simulation program MODFLOW2000
(Harbaugh et al. 2000) and the multi-species mass transport
simulation program MT3DMS (Zheng and Wang 1999), which
enables simulation of variable-density groundwater flow. To al-
low for heat transport, the differential equations for solute trans-
port in SEAWATv4 were translated in terms of heat transport
following the approach described by Langevin et al. (2008).
Equations of state for density and viscosity were used to
describe both fluid density and viscosity as a function of tem-
perature and salt concentration. Fluid viscosity as a function of
temperature and salt concentration is described by the follow-
ing equation (Voss 1984):
μCS;TðÞ¼2:394⋅10−5⋅10 248:37
Tþ133:15
ðÞ
þ1:923⋅10−6CS
ðÞ ð1Þ
where μis the dynamic fluid viscosity (kg/m day), Tis the
temperature of the water (°C) and C
s
is the solute concentra-
tion of the water (kg/m
3
).
A non-linear density equation of state derived by Sharqawy et
al. (2010) was used in the SEAWAT code as described by Van
Lopik et al. (2015) to accurately simulate the temperature-density
relation over large temperature ranges (see Eq. 2). This density
relationship based on experimentally derived datasets for both salt
concentration and temperature at 1 atm pressure from Isdale and
Morris (1972) and Millero and Poisson (1981)isshowninFig.2.
ρT;SðÞ¼999:9þ2:034⋅10−2T−6:162⋅10−3T2þ2:261⋅10−5T3−4:657⋅10−8T4
þ
802:0S
1000 −2:001 S
1000 Tþ1:677⋅10−2S
1000 T2−3:060⋅10−5S
1000 T3−1:613⋅10−5S
1000
2
T2
! ð2Þ
where ρis the fluid density (kg/m
3
)andSis the salinity of the
water (g/kg).
Hydrogeol J
Model set-up
The modeling in the present study is done in two parts. Firstly, a
sensitivity analysis is performed by simulation of seasonal HT-
ATES with four consecutive recovery cycles (see section
‘Generalized HT-ATES cases used for the sensitivity analysis’).
Secondly, the density difference compensation method was
tested for an actual HT-ATES field pilot of two recovery cycles
in a high-permeability aquifer conducted by Molz et al. 1983a
(see section ‘Pilot study at the Auburn University’). For both
the sensitivity analysis on seasonal HT-ATES and the numerical
simulation of the field pilot, a confined sandy aquifer was con-
Fig. 1 Schematic overview of a
full HT-ATES recovery cycle in a
confined aquifer for ainjection, b
storage and cextraction periods.
Heat loss occurs by thermal
conduction and free thermal
convection
Fig. 2 Water density as a
function of temperature and
salinity (Sharqawy et al. 2010,
Eq. 2). Solid lines represent equal
water densities (isopycnals). The
black stars show the required
salinity for HT-ATES water
injection at temperatures of 60
and 80 °C into brackish
groundwater with a temperature
of 20 °C (brackish = 10 g/kg). The
red cro sses are for water injection
at temperatures of 58.5 and 81 °C
into fresh groundwater with a
temperature of 20 °C
Hydrogeol J
sidered to simulate HT-ATES recovery cycles at the various
temperature and salt-concentration contrasts between injected
water and ambient groundwater. Groundwater flow and associ-
ated heat and salt transport were simulated axi-symmetrically,
following the approach introduced by Langevin (2008). This
approach has been validated for transport of solute (Wallis et al.
2013) and heat (Vandenbohede et al. 2014). Although the dis-
placement of heat and solutes by regional groundwater flow
cannot be considered with the axi-symmetric approach, the re-
gional groundwater flow is generally low in deep brackish aqui-
fers targeted for HT-ATES systems (e.g. Sauty et al. 1982).
Hence, neglecting heat and solute loss by regional groundwater
flow during HT-ATES is considered a reasonable assumption.
The axi-symmetric model domain has a radius of 500 m
with an aquifer thickness of 21 m. The overlying and under-
lying aquitards are 10 and 9 m thick. The grid resolution is
Δr=0.5 m by Δz= 0.5 m. Constant head, temperature and
concentration boundaries were used for the outer, upper and
lower boundaries of the model domain, following Buscheck
et al. (1983). The inner boundary is impermeable inside the
aquitards and simulates the well inside the aquifer. The well
used for both injection and extraction has a radius of 0.1 m.
The groundwater flow is solved using the Preconditioned
Conjugate Gradient 2 (PCG2) package. The modified method
of characteristics (MMOC) is applied as an advection package
with a Courant number 0.2. In order to simulate heat conduc-
tion accurately, the convergence criterion of relative tempera-
ture is set to 10
−10
°C (Vandenbohede et al. 2014).
Generalized HT-ATES cases used for the sensitivity analysis
The aquifer characteristics used for the simulation of seasonal
HT-ATES are based on the aquifer used for the Auburn
University (USA) field experiment conducted by Molz et al.
(1983a). A homogeneous anisotropic aquifer has been used for
the sensitivity analysis instead of the heterogeneous layering
described by Buscheck et al. (1983). The aquifer characteristics
for this base of the sensitivity analysis (case 1) are listed in
Tab le 1. The ambient groundwater in the sensitivity analysis
has a brackish salinity of 10,000 ppm (C
s
=10 kg/m
3
).
A seasonal HT-ATES system was assumed, with injection,
storage, extraction and rest periods of 90 days each. Four con-
secutive cycles were simulated to investigate how the thermal
recovery efficiency develops with time. An equal injection and
extraction volume of 56,700 m
3
andaninjectiontemperatureof
80 °C were assumed for each cycle in the reference scenario
(case 1,Table2). Three types of simulations (named a, b and c)
were conducted for all cases in the sensitivity analysis: regular
HT-ATES (e.g. case 1.a for the reference scenario), HT-ATES
with density difference compensation using saline water (e.g.
case 1.b for the reference scenario), and a theoretical regular
HT-ATES case that considers no free thermal convection and,
hence, only heat loss by thermal conduction (e.g. case 1.c for the
reference scenario). In the latter simulation (c), density-driven
flow was not considered (no free thermal convection) as an
upper bound for the improvement that can be obtained with
density difference compensation. In the additional five cases
with each three types of simulations, the sensitivity of the model-
ing results was investigated by varying HT-ATES conditions
and aquifer characteristics (Table 2).
Pilot study at the Auburn University
The simulation (Buscheck et al. 1983) and experimental re-
sults (Molz et al. 1983a) of the HT-ATES field experiment
conducted at the Auburn University were used to test the
Tabl e 1 Aquifer and aquitard properties of the Auburn University field
experiment (Molz et al. 1983a)
Properties Parameter values
Aquifer properties S
s
=6·10
−4
m
−1
Specific storage S
s
=6·10
−4
m
−1
Porosity θ=0.25
Bulk density ρ
b
= 1,950 kg/m
3
Heat capacity c
ps
=696.15 J/kg °C
Thermal conductivity of aquifer λ
s
= 2.29 W/m °C
Thermal distribution coefficient
a
K
dT
=1.66·10
−4
m
3
/kg
Thermal retardation factor
a
R
T
=2.29
Bulk thermal diffusivity
a
D
T
=0.189 m
2
/day
Overall horizontal hydraulic conductivity k
h
= 53.4 m/day
Overall vertical hydraulic conductivity k
v
= 7.7 m/day
Aquifer thickness H
a
=21 m
Aquitard properties
Specific storage S
s
=9·10
−2
m
−1
Porosity θ=0.35
Bulk density ρ
b
= 1,690 kg/m
3
Heat capacity c
ps
=696.15 J/kg °C
Thermal conductivity λ
s
= 2.56 W/m °C
Thermal distribution coefficient
a
K
dT
=1.66·10
−4
m
3
/kg
Bulk thermal diffusivity
a
D
T
=0.151m
2
/day
Horizontal hydraulic conductivity k
h
= 53.4 m/day
Vertical hydraulic conductivity k
v
= 7.7 m/day
Groundwater properties
Heat capacity of the fluid c
pf
= 4,186 J/kg °C
Thermal conductivity of the fluid λ
l
=0.58 W/m °C
Solute transport properties
Longitudinal dispersion α
l
=0.5 m
Transversal dispersion α
t
=0.05 m
Molecular diffusion D
m
=8.64·10
−5
m
2
/day
a
The thermal distribution coefficient (K
dT
), thermal retardation factor
(R
T
), bulk thermal diffusivity (D
T
) are calculated for SEAWATv4 heat
transport simulation (see Langevin et al. 2008)
Hydrogeol J
potential of density difference compensation with respect to
the regular HT-ATES operated in this specific field experi-
ment. The main aspects from Buscheck et al. (1983)and
Molz et al. (1983a) are summarized here.
Two injection-storage-recovery cycles (injection tempera-
tures of 58.5 and 81 °C) were conducted in a highly permeable
sandy confined aquifer. The first cycle was of 3 months dura-
tion, while the second cycle had a duration of 7.3 months. The
determined average production and injection volume rates and
injection temperatures over time as presented by Buscheck et
al. (1983) were used in the simulations of this study (Table 3).
The characteristics of the aquifer and the confining layers used
for the numerical modeling of this field experiment are listed
in Table 1. A three-layered heterogeneous aquifer was consid-
ered with a 2.5-fold higher hydraulic conductivity for the mid-
dle layer than for the upper and lower layers (Table 4)inorder
to predict the two recovery cycles numerically (Buscheck
et al. 1983). The ambient groundwater temperature is 20 °C
and contains only 280 ppm or 0.28 kg/m
3
of total dissolved
solids (Molz et al. 1983b). Due to low recovery temperatures
of extracted water during the second recovery cycle, the well
configuration was changed. The original fully penetrating well
screen was replaced by a partially penetrating well screen after
2 weeks of extraction to increase thermal energy recovery,
resulting in a recovery efficiency of 0.45 (Molz et al. 1983a).
Molz et al. (1983a) estimated a thermal energy recovery effi-
ciency of 0.40 without the partially penetrating well screen mod-
ification. Buscheck et al. (1983) simulated the second recovery
cycle for a fully penetrating well screen. The same procedure is
followed in the numerical simulation with SEAWATv4.
Density difference compensation using saline water
for heat storage
The required salinity for the density difference compensation
was calculated for the heat injection temperature based on the
non-linear density equation (Eq. 2)asillustratedinFig.2. Salt
mass transport was modeled conservatively and adsorption or
precipitation reactions were not incorporated in the numerical
simulations.
In the sensitivity analysis for various HT-ATES conditions
(see section ‘Generalized HT-ATES cases used for the sensi-
tivity analysis’), a brackish salinity of 10,000 ppm (10 kg/m
3
)
was selected for the ambient groundwater, as environmental
considerations typically preclude the use of HT-ATES in fresh
Table 2 Summary of the input parameters used for the sensitivity
analysis
k
h
[m/day] k
v
[m/day] T
in
[°C] H
a
[m] V
in
[m
3
]
Case 1 53.4 7.7 80 21 56,700
Case 2 15
a
1.5
a
80 21 56,700
Case 3 53.4 7.7 60
a
21 56,700
Case 4 53.4 7.7 80 10
a
56,700
Case 5 53.4 7.7 80 21 28,350
a
Case 6 53.4 7.7 80 21 113,400
a
a
Indicates a variation on the reference scenario (case 1)
Tabl e 3 Injection and extraction
flow rates, as well as injection
temperatures during the first and
second cycle of the Auburn
University field pilot (Buscheck
et al. 1983). These values are used
for the numerical simulation in
this study
Cycle Phase Length [days] Volumetric flow rate [m
3
/day] Injection temperature [°C]
Cycle-1
Injection 20 760 60
7 1,100 58
4600 52
Storage 32 - -
Extraction 2 −1,684.8 -
21 −1,054.1 -
Rest 26 - -
Cycle-2
Injection 8 954 85
24 - -
7600 82
4- -
85 545 80
Storage 34 - -
Extraction 14 −1,088.6 -
2- -
39 −1,088.6 -
Hydrogeol J
groundwater aquifers. The salt concentration required to es-
tablish the same water density as ambient groundwater
(T=20 °C and C
s
=10 kg/m
3
) for hot injection water at tem-
peratures of 60 and 80 °C is shown in Fig. 2. The required salt
concentrations are 30.7 and 46.5 kg/m
3
,respectively.
For the HT-ATES pilot study at Auburn University
(see section ‘Pilot study at the Auburn University’),
the density difference between hot injection water and
ambient groundwater was calculated for both cycles
with injection temperatures of 58.5 and 81 °C, respec-
tively. Consequently, the required salt concentration to
overcome buoyancy difference between injected water
and ambient groundwater was 19 kg/m
3
for cycle-1,
whereas for cycle-2, a salt concentration of 36.6 kg/m
3
was required (Fig. 2).
Metrics to quantify salt and heat recovery
Critical in the operation of HT-ATES systems is how
much of the injected heat can be recovered following
storage, expressed as thermal recovery efficiency. The
thermal recovery efficiency (ε
H
) is defined as the ratio
between the total injected heat (Q
in
) and the total recov-
ered heat after extraction (Q
ex
). The total injected and
recovered heat is calculated from the temperatures of
injected and extracted water:
εH¼Qex
Qin
¼XVex⋅ρex T;SðÞ⋅cpf⋅Tex−Ta
ðÞ
XVin⋅ρin T;SðÞ⋅cpf⋅Tin−Ta
ðÞ
ð3Þ
where V
ex/in
are the volumes per time step of the extracted
water and the injected water (m
3
), T
ex/in
are the temperatures
of the extracted water and the injected water (°C), T
a
the am-
bient groundwater temperature (°C), c
pf
the heat capacity of
water, J/(kg °C), and ρ
in/ex
(T,S) the densities of the extracted
water and the injected water as a function of salt concentration
and temperature (Eq. 2). In the sensitivity analysis (cases1–6),
summation is done over the output time steps for salinity and
temperature (Δt= 5 days) to calculate Q
in
and Q
ex
. For the
Auburn University HT-ATES field experiment, smaller output
time steps of 1 day are used.
For density difference compensation, the recovery of salt
mass (ε
S
) is an important variable in this study. The salt mass
recovery efficiency is defined as the ratio between the total
injected salt mass (M
Sin
) and the total recovered salt mass
after extraction (M
Sex
):
εS¼MSex
MSin
¼XVex⋅CSex
−CSa
ðÞ½
XVin⋅CSin
−CSa
ðÞ½ ð4Þ
where C
Sex/in
are the salt concentrations in each time step of
the extracted water and the injected water (kg/m
3
)andC
Sa
is
the ambient groundwater concentration (kg/m
3
).
Spreading of hot and saline water
In the sensitivity analysis, theoretical optimal HT-ATES cases
that considered only heat loss by thermal conduction (e.g. case
1.c) are run and compared to density difference compensation
(e.g. case 1.b). Considering an optimal case with no heat loss,
the shape of the stored hot water in the aquifer will be cylin-
drical. Therefore, the dimensions of the hot water volume can
be described in terms of the maximum radial extent of the
injected hot water (r
th
) and a given aquifer thickness (H
a
).
Equilibration of heat between the solid and water phase
occurs during heat transport in the aquifer. Therefore, the tem-
perature front progresses at a lower rate into the aquifer than
the effective flow velocity during hot water injection. The
thermal radius of the hot injection water (r
th
) is defined by
the cylinder formed by retarded advective transport only:
rth ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Vin
πHaθRT
rð5Þ
where θis the porosity, H
a
is confined aquifer thickness (m)
and R
T
is the thermal retardation factor:
RT¼1þρb
θ
cps
ρfcpf
ð6Þ
where ρ
b
is the bulk density (kg/m
3
)andc
ps
the heat capacity
of solid, J/(kg °C).
Similar to the thermal radius, a solute radius (r
s
)canbe
defined considering conservative advective solute transport
and neglecting molecular diffusion and dispersion:
rs¼ffiffiffiffiffiffiffiffiffiffiffiffi
Vin
πHaθ
rð7Þ
Results
The results of this study are presented in two parts. First, the
sensitivity analysis are presented in which various seasonal
Tab l e 4 Hydraulic conductivity of the heterogeneous aquifer at
Auburn University, according to Buscheck et al. (1983)
Hydraulic conductivity Value
Horizontal hydraulic conductivity
Upper layer (10 m thick) k
h
=38.2 m/day
Middle layer (5 m thick) k
h
=96.4 m/day
Lower layer (6 m thick) k
h
=38.2 m/day
Vertical to horizontal hydraulic conductivity ratio 1:7
Hydrogeol J
HT-ATES cases have been simulated for three scenarios: (a) a
regular setup, (b) a setup with density difference compensa-
tion using saline water, and (c) a setup under the theoretical
condition of no free thermal convection. Secondly, the results
of the SEAWATv4 modelling of the two recovery cycles of the
Auburn University field experiment performed with and with-
out density difference compensation are presented.
Sensitivity analysis
Several cases were simulated to test the potential of density
difference compensation for HT-ATES and its sensitivity for
various conditions (see section ‘Generalized HT-ATES cases
used for the sensitivity analysis’). Four parameters were var-
ied with respect to the reference scenario (case 1, Table 2): a
lower aquifer hydraulic conductivity (case 2), a lower injec-
tion temperature (case 3), a smaller aquifer thickness (case 4),
and a smaller and larger injection volume (cases 5 and 6).
Reference scenario (case 1)
In case 1, a seasonal HT-ATES system with injection temper-
atures of 80 °C is simulated both with and without density
difference compensation. Without density difference compen-
sation, the results showed strong free thermal convection dur-
ing injection and storage for regular HT-ATES (case 1.a,
Fig. 3). Consequently, unheated ambient groundwater is re-
covered at the lower part of the well screen and overall recov-
ered water temperatures over time ranged between 49.6 and
33.5 °C for cycle-1 (Fig. 4). The thermal recovery is low,
ranging from 0.31 to 0.40 for the four recovery cycles
(Fig. 5). Initially, the increase in thermal recovery efficiency
is large (Δε
H
= 0.051), but for the subsequent cycles the in-
crease in recovery efficiency gradually decreases to an Δε
H
increase of only 0.015 from cycle-3 to cycle-4. The buoyancy
of the injected hot water decreases for each consecutive recov-
ery cycle, which results in less pronounced thermal convec-
tion. Also the heat loss by thermal conduction decreases due
to a lower temperature difference between injection water and
ambient groundwater.
For the scenario with density difference compensation
(case 1.b), the required salt concentration to reduce the density
contrast between the hot injection water and the ambient
groundwater is 46.5 kg/m
3
(Fig. 2). This means that a total
salt mass of 2.07 · 10
6
kg was injected for each recovery cycle.
The results show that free thermal convection during injection
and storage was successfully countered (Fig. 6). A large in-
crease in thermal recovery (Δε
H
= 0.29) was obtained with
respect to the regular HT-ATES (case 1.a) for each recovery
cycle (Fig. 5). Moreover, the recovery temperatures over time
were significantly better preserved with respect to the injec-
tion temperature (Fig. 4). The production temperature slightly
decreases during the first 50 days of extraction when the sta-
bilized hot water volume is recovered. This resulted in in-
creased recovery temperatures up to 20 °C with respect to
the regular HT-ATES (case 1.a). After 50 days, a rapid decline
in production temperature is observed during recovery. In this
period, hot water was recovered which was largely affected by
Fig. 3 Temperature distribution
for the reference scenario of
regular HT-ATES without density
difference compensation (case
1.a): aat the end of the injection
period, bat the end of the storage
period and cat the end of the
extraction period
Hydrogeol J
the radial heat loss due to thermal conduction into the aquifer
in the preliminary injection and storage stages, as was by the
similar temperature decline when modeling the scenario with-
out accounting for free thermal convection (case 1.c). Thus,
besides reducing the heat loss due to free thermal convection,
applying density difference compensation also reduced the
heat loss by thermal conduction in comparison to the regular
HT-ATES. This is due to the more compact, cylindrical hot
water volume in the aquifer that is maintained when applying
density difference compensation, which resulted in a smaller
total contact area of both the colder surrounding aquifer and
colder confining layers where heat is lost by thermal conduc-
tion (Fig. 6). The difference in thermal recovery between den-
sity difference compensation (case 1.b) and the theoretical
case with no free thermal convection (case 1.c) is small
(Δε
H
= 0.07). This means that for case 1.b, the main heat loss
occurs by thermal conduction. However, as reflected by the
initially lower recovery temperatures compared to the scenario
that ignores free thermal convection (case 1.c) for the first
50 days of extraction (Fig. 4), a small fraction of heat loss still
occurs by a different mechanism. This is due to the fact that in
the density difference compensated HT-ATES system free
convection is determined by the density contrasts resulting
from both the salt concentration and temperature differences
in the aquifer. The thermal retardation (Eq. 6) affects heat
transport, resulting in faster advective solute transport than
heat transport during injection operations (e.g. Miotliński
and Dillon 2015; Seibert et al. 2014). Therefore, plume sepa-
ration of hot water (see Eq. 5) and saline water (see Eq. 7)
occurs duringHT-ATES with density difference compensation
(case 1.b), which results in a cold saline water plume in front
of the hot water volume. Oldenburg and Preuss (1999)
showed that plume separation of saline and hot water highly
affects convective flow patterns if the salt concentration and
the temperature contrasts are large. For HT-ATES with density
difference compensation (case 1.b), the injected salt
concentration of 46.5 kg/m
3
creates a strong density contrast
at ambient temperature with the ambient groundwater
concentration (C
s
=10 kg/m
3
) of approximately 28 kg/m
3
(1,034 kg/m
3
vs. 1,006 kg/m
3
, Fig. 2). Hence, a downward
buoyancy force is exerted on the cooled saltwater front
resulting in downward transport (Fig. 6d–e). Due to down-
ward density-driven flow of cold saltwater, the hot water vol-
ume at the bottom of the aquifer is displaced towards the well
screen. Therefore, cold saline water was recovered at the low-
er part of the well screen (Fig 6b), resulting in lower recovery
efficiencies than the theoretical optimal case when not consid-
ering free thermal convection (case 1.c). In the last stage of the
extraction period (after 260 days), the recovered temperature
is slightly higher for case 1.b than for case 1.c (Fig. 4). The
displacement of the cold saline water towards the well screen
Fig. 4 Production temperatures over time for the first recovery period of
case 1
Fig. 5 Calculated athermal and
bsalt mass recovery efficiency
per cycle for the reference case 1
Hydrogeol J
resulted in lateral transport of heated water away from the well
screen in the upper part of the aquifer. Consequently, the heat-
ed water volume in the upper part of the well screen has
moved further from the well than in the no free thermal
convection scenario (case 1.c). In the final stage of extraction,
this displaced heated water volume at the top of the aquifer is
recovered, resulting in slightly higher recovery temperatures
for the density difference compensation method.
In addition to the thermal recovery, the salt recovery
was determined for the density difference compensation
method, at 0.54 for the first cycle, after which it increases
to a value of 0.68 in the fourth cycle (Fig. 5b). This
resulted in an unrecovered salt mass of 1.21· 10
6
and
8.42 · 10
5
kg, respectively. Salt mass accumulates at the
bottom of the aquifer after each extraction period. Only
a small amount of saltwater is accumulated in the upper
clay layer due to molecular diffusion during the injection
and storage phase (Fig. 6f). For the first cycle, saltwater
(C
s
> 12 kg/m
3
) is transported up to a radial distance of
124 m along the bottom of the aquifer, while this is 201 m
for cycle-4.
Hydraulic conductivity of the aquifer (case 2)
Currently, low-permeability aquifers are selected for HT-ATES
systems to reduce free thermal convection and to obtain suffi-
ciently high thermal recovery efficiencies (e.g. Doughty et al.
1982;Schoutetal.2014). For case 2, an aquifer with a hori-
zontal and vertical hydraulic conductivity of 15 and 1.5 m/day,
respectively, was considered to account for such aquifers char-
acteristics (instead of 53.4 and 7.7 m/day for case 1). The ther-
mal recovery efficiencies for the regular HT-ATES (case 2.a)
Fig. 6 Temperature distribution
for the reference case with density
difference compensation (case
1.b) at the end of athe injection
period, bof the storage period,
and cof the extraction period.
Black contour lines indicate the
temperature distribution for the
regular reference case 1.a. The
same is done for the salt
concentration distribution at the
end of dthe injection period, eof
the storage period, and fof the
extraction period
Hydrogeol J
are elevated compared to the reference scenario (case 1.a) with
values of 0.58 and 0.64, for the first cycle and fourth cycle
(Fig.7a), respectively. Although the effect was less, density dif-
ference compensation (case 2.b) still improved thermal recov-
ery efficiency (Δε
H
) with 0.10 compared to regular HT-ATES
(case 2.a). The recovered temperature declines substantially
during extraction (Fig. 8). The downward salt-water transport
between the solute and thermal front causes less displacement
of the hot water towards the well screen due to decreased flow
velocities. Therefore, the difference in thermal recovery (Δε
H
)
between density difference compensation (case 2.b) and the
optimal case with no free thermal convection (case 2.c) is only
0.02 for cycle-4 (Fig. 7a). Despite the significantly lower per-
meability, the efficiency of regular HT-ATES (case 2.a) was
lower than for the density difference compensation in the
high-permeability reference scenario (case 1.b).
Salt mass recovery efficiency is much higher for case 2.b,
compared to the reference scenario, case 1.b (Fig. 7b).
Downward density-driven flow of cold saltwater is restricted
due to decreased hydraulic conductivity of the aquifer.
Consequently, the accumulation rate of salt mass at the bottom
of the aquifer is low compared to case 1.b.
Fig. 7 a Calculated thermal and
bsalt recovery efficiency of
cycle-4 for all cases
Fig. 8 Production temperatures over time of cycle-1 for case 2
Hydrogeol J
Injection temperature (case 3)
An injection temperature of 60 °C was considered for case 3, to
test the effect of a lower temperature difference between hot
injection water and ambient groundwater (Table 2). For regular
HT-ATES, a higher thermal recovery efficiency was achieved
with a lower injection temperature (case 3.a; T
i
=60°C)thanfor
the reference scenario (case 1.a; T
i
= 80 °C), due to less pro-
nounced free thermal convection. For the theoretical cases
where no free thermal convection was considered (case 1.c
and case 3.c), the calculated thermal recovery efficiency was
0.76 for both. This independence of recovery efficiency on the
injection temperature is due to the linear increase of both heat
loss by thermal conduction and the total injected heat (Q
in
),
with the resulting temperature contrast between hot injection
water and the cold surrounding aquifer and aquitards.
For the density difference compensation scenario, a salt
concentration of 30.7 kg/m
3
(Fig. 2) was required to overcome
the buoyancy difference between hot injection water and am-
bient groundwater. The HT-ATES scenario with density dif-
ference compensation (case 3.b), resulted in a net recovery
efficiency increase (Δε
H
) of 0.18, with respect to regular
HT-ATES (case 3.a). The recovery efficiency was close to
the theoretical case with no free thermal convection (case
3.c) with a Δε
H
value of only 0.05 (Fig. 7a). This is caused
by the less-pronounced downward saltwater transport for this
scenario than in the reference scenario (case 1.b), due to the
lower salt concentration contrast with the ambient groundwa-
ter. Consequently, the induced hot water displacement towards
the well screen is less pronounced. In addition, the calculated
salt recovery efficiency is higher for case 3.b than for the
reference scenario (case 1.b), since the accumulation rate of
salt at the bottom of the aquifer is lowered (Fig. 7b).
Aquifer thickness and injection volume (cases 4–6)
The sensitivity of the recovery efficiency for regular HT-
ATES and with density difference compensation was tested
for different injection volumesand aquifer thickness (Table 2).
This was done simulating a two times smaller aquifer thick-
ness (case 4), as well as a two times smaller (case 5) and a two
times larger injection volume (case 6) with respect to the ref-
erence (case 1).
For the regular HT-ATES cases 4.a and 6.a, the recovery
efficiencies are higher and the maximum radii of heated vol-
ume are larger (r
th
, Table 5). The contrary is observed for the
case with the smaller volume (case 5.a) which has the smallest
r
th
value. For regular HT-ATES in a high-permeability aquifer,
heat loss is mainly determined by free thermal convection.
Consequently, thermal front tilting close to the well screen
(e.g. case 5.a) resulted in recovery of large amounts of cool
ambient groundwater at the lower part of the well screen dur-
ing the extraction stage. Therefore, the largest increase in ther-
mal recovery efficiency with density difference compensation
is obtained for the scenarios with a small r
th
value, resulting in
aΔε
H
value of 0.28 for case 1.b (compared to 1.a) and 0.33
for case 5.b (compared to 5.a), respectively. A smaller aquifer
thickness (case 4.b) results in an increase of only 0.04, while
for a higher injection volume (case 6.b) this is 0.20 (Fig. 7a).
The differences in thermal recovery between density differ-
ence compensation and the theoretical scenario without free
thermal convection are small, with Δε
H
ranging from 0.06 to
0.08 for cases 1 and 4–6(Fig.7a). The hot water volume is not
completely stabilized with density difference compensation,
since hot water is displaced towards the well screen due to
downward transport of the cold saltwater behind the solute front.
However, this effect on thermal recovery is small and heat loss is
mainly due to thermal conduction. Table 5shows that the
highest energy recovery is obtained for case 6.c, where the ratio
between the outside area and thermal volume in the aquifer is
lowest (0.13 m
−1
). A smaller injection volume (case 5.c) results
in a larger ratio of 0.17 m
−1
and a lower recovery efficiency. For
small injection volumes, the overall surface area will be relative-
ly large compared to the thermal volume. Therefore, the heat
losses due to thermal conduction will be high.
The relative conductive heat loss to the surroundings with
respect to the total amount of stored heat is highest for cases
4.b–c, with an area volume ratio of 0.24 m
−1
. Hence, the
Tabl e 5 Overview of the
sensitivity analysis on thermal
and salt recovery efficiency for
cycle-4. The associated thermal
radii (r
th
,Eq.5), the solute radii
(r
s
,Eq.7), and the ratio between
the outside area and the thermal
volume of the theoretical cylinder
in the aquifer (A
tot
/V
H
)
Scenario r
th
[m] A
tot
/V
H
[m
−1
] Thermal recovery efficiency r
s
[m] Salt recovery
efficiency
Dens. diff.
compensation
No free
convection
Case 1
reference
38.75 0.15 0.69 0.76 58.63 0.68
Case 4
H
a
=10 m
56.15 0.24 0.61 0.67 84.97 0.78
Case 5
V
i
=28,350 m
3
27.40 0.17 0.65 0.73 41.46 0.59
Case 6
V
i
=113,400 m
3
54.79 0.13 0.70 0.78 82.99 0.74
Hydrogeol J
lowest thermal recovery (0.61) was obtained with the density
difference compensation method (case 4.b) compared to all
other scenarios.
Salt mass recovery efficiency is smallest for the scenarios
with the cold saltwater front close to the well screen (e.g. case
5.b; Table 5and Fig. 7b). Although the injection volume in
case 5.b is only half of the reference scenario (case 1.b), salt-
water (C
s
> 12 kg/m
3
) is already transported up to a radial
distance of 105 m along the bottom of the aquifer for case
5.b for the first cycle, while this is 124 m for case 1.b. This
means that the relative salt displacement away from the well
screen is larger for cases with a cold saltwater front closer to
the well screen. For scenarios with a larger radius of injected
saltwater (r
s
), the cold saltwater can be transported over a
larger distance towards the well screen and hence the relative
salt mass loss due to lateral transport away from the well
screen is lowered (e.g. cases 4.b and 6.b; Table 5).
Modeling of the pilot study at Auburn University
The numerical SEAWATv4 modeling results of the two
injection-storage-extraction cycles as conducted in the
Auburn University field pilot (Molz et al. 1983a)
corresponded well with the modeling results by Buscheck et
al. (1983). They simulated the two cycles of the field experi-
ment with the computer program PT using a non-linear tem-
perature-density relationship (Bodvarsson 1982).
Figure 9shows the calculated temperature distributions at
the end of the first injection period with an injection temper-
ature of 58.5 °C. The observed preferential flow in the middle
high-permeability layer is reproduced. Similar to the modeling
results of Buscheck et al. (1983), the higher injection temper-
ature (81 °C) in the second cycle led to stronger free thermal
convection resulting in no discernible preferential flow due to
the heterogeneous layering (Fig. 10). The thermal recovery
efficiency calculated for the simulations of the two cy-
cles was 0.58 and 0.41, respectively, similar to both the
experimental results of Molz et al. (1983a) and the nu-
merical calculations by Buscheck et al. (1983), as
shown in Table 6. For the second cycle, a smaller tran-
sition zone between the hot water volume and ambient
groundwater was obtained in this study compared to the
results of Buscheck et al. (1983;seeFig.10). Moreover,
enhanced free thermal convection is observed in the
SEAWAT modeling results of this study. These differ-
ences are likely explained by the smaller grid sizes used
in the thermal zone in the SEAWAT model (Δr=0.5 m)
than in the model (Δr=4.0 m) of Buscheck et al.
(1983), causing less numerical dispersion for heat
transport.
Simulating density difference compensation for the Auburn
University pilot study
The potential of the density difference compensation method
was tested using the model that adequately modeled the
Auburn University field pilot. The required salt concentrations
to overcome the buoyancy difference between hot injection
water (58.5 °C for cycle-1 and 81 °C for cycle-2) and the
ambient groundwater (20 °C) are 19 and 36.6 kg/m
3
,respec-
tively (Fig. 2). Free thermal convection was successfully
countered for the second cycle (Fig. 11), so that the preferen-
tial flow of the hot, saline water in the most permeable layer is
more pronounced (compare Fig. 11 with Fig. 10). The prefer-
ential flow of saline water into the middle layer of the hetero-
geneous aquifer results in injected cold saltwater overlying
cold fresh ambient groundwater in the bottom layer. The sa-
linity contrast of 36.6 kg/m
3
results in a density difference
between ambient fresh water (998.0 kg/m
3
)andthecoldsalt-
water front (1,026.1 kg/m
3
). Therefore, downward transport
Fig. 9 The color intensity
indicates the temperature
distribution calculated by
SEAWAT at the end of the
injection period for cycle-1. Black
contour lines indicate the
temperature distribution modeled
by Buscheck et al. (1983)
Hydrogeol J
of cold saltwater mass is observed, which results in more
accumulation of salt mass at the bottom of the aquifer over
time in this heterogeneous aquifer than in the homogeneous
aquifer used in the sensitivity analysis. Consequently, in-
creased displacement of the hot water volume at the bottom
of the aquifer towards the well screen is observed. For both
cycles, density difference compensation increased the thermal
recovery efficiency to 0.68 and 0.66 for cycle 1 and 2 respec-
tively (Table 7). The relative increase in thermal recovery
efficiency was large (60.8 %) for cycle-2, which had an effi-
ciency of 0.41 in the regular HT-ATES. The salt recovery
efficiency was 0.69 for cycle-1 and 0.63 for cycle-2.
Discussion
In this study is shown that free thermal convection during HT-
ATES can be countered by using saline water for heat storage.
This density modification allows a significant increase of re-
covery efficiency, bringing it close to the theoretical situation
without free thermal convection. This has the advantage that
HT-ATES systems are no longer restricted to low-permeability
aquifers to achieve sufficient heat recoveries. Also, larger tem-
perature differences between the hot water volume and cold
ambient groundwater can be used, which increases the capac-
ity and efficiency of the building systems.
Optimization of thermal recovery efficiency
The recovery efficiency of regular HT-ATES systems is sig-
nificantly influenced by both thermal conduction and free
thermal convection. The thermal front tilting due to free ther-
mal convection causes unheated ambient groundwater to be
recovered at the lower part of the well screen, especially for
cases with a small thermal radius (r
th
, case 5.a). Therefore,
feasibility of regular HT-ATES is limited (Doughty et al.
1982;Schoutetal.2014). In the past, several suggestions have
been put forward to optimize energy recovery during HT-
ATES. In order to reduce the heat loss by free thermal con-
vection during injection and storage, partially penetrating
wells can be used to increase energy recovery (Buscheck
et al. 1983). However, only a small increase in recovery effi-
ciency was obtained (Δε
H
< 0.09). Most r esearch on optimiz-
ing well designs of partially penetrating well systems is done
in the field of aquifer storage and recovery (ASR), where
density-driven flow occurs due to the salinity contrast between
injected freshwater and brackish groundwater (Ward et al.
2007). For these cases, well systems with multiple injection
and recovery wells (Miotliński et al. 2014)ormultiplepartial-
ly penetrating wells can be used to successfully increase the
recovery of freshwater (e.g. Zuurbier et al. 2014). Also
prolonged injection of hot water during the first cycle can be
applied to reduce the temperature differences and associated
heat loss in subsequent recovery cycles (Sauty et al. 1982).
Fig. 10 The color intensity
indicates the temperature
distribution calculated by
SEAWAT at the end of the
injection period for cycle-2. Black
contour lines indicate the
temperature distribution
calculated by Buscheck et al.
(1983)
Tabl e 6 Comparison between
experimental and calculated
recovery efficiencies
Experimental
(Molz et al. 1983a)
Numerical estimate
(Buscheck et al. 1983)
Numerical estimate
(SEAWATv4, this study)
Cycle-1 0.56 0.58 0.58
Cycle-2 0.40
a
0.40 0.41
a
Estimated value by Molz et al. 1983a assuming no well screen modification during production
Hydrogeol J
This study showed that the thermal recovery efficien-
cy of HT-ATES systems can be efficiently increased by
applying density difference compensation using saline
water. Consequently, heat loss due to thermal conduc-
tion remains as the main factor affecting thermal recov-
ery efficiency, which enables the optimization of HT-
ATES heat recovery by minimizing conductive heat
loss, rather than mitigating heat losses by free thermal
convection. The conductive heat loss is related to the
compactness of the heated volume. Relatively large out-
side areas of the injected hot water volume result in a
strong increase of heat loss by thermal conduction (e.g.
case 4.b).
Besides the heat loss due to thermal conduction, the appli-
cation of the density difference compensation method also
induces displacement of hot water volume towards the well
along the bottom of the aquifer due to downward density-
driven flow of the cold saltwater between the solute front
and the thermal front (Fig. 6). Although this only has a slight
negative impact on the thermal recovery efficiency, these
double-advective effects on heat and salt transport during
HT-ATES need to be taken into account to fully optimize the
thermal and salt recovery efficiency, especially for heteroge-
neous aquifers where the effect of plume separation of the
heated and saline water volume could have a large impact
on convective flow patterns due to the density contrasts (see
section ‘Simulating density difference compensation for the
Auburn University pilot study’).
The calculated recovery efficiency for the simulation of
seasonal HT-ATES in this study is determined by assuming
equal injection and production volumes. In practice, the ex-
traction period may be reduced by maintaining a minimum
temperature for extraction. The calculated production temper-
ature over time is significantly higher for HT-ATES with den-
sity difference compensation, compared to regular HT-ATES
(Figs. 4and 8). Therefore, the increase in recovery efficiency
while using density difference compensation is larger when
applying a minimum extraction temperature. For example, the
thermal recovery efficiency increase with density difference
compensation is relatively small for a low-permeability aqui-
fer (Fig. 7a, case 2.b) while using equal injection and produc-
tion volumes. However, if a minimum temperature of 60 °C is
used, the extractable volume at a constant extraction rate of
630 m
3
/day for regular HT-ATES (case 2.a) was only 17,
010 m
3
, whereas this is 30,870 m
3
using density difference
compensation (case 2.b).
Model validation on the Auburn University field
experiments
Recently, Vandenbohede et al. (2014) validated the approach
of Langevin (2008) for heat transport on an ATES test case,
where water with a temperature of 17.5 °C was injected into
an aquifer with an ambient groundwater temperature of 10 °C.
Due to this small temperature contrast, no density-driven flow
was considered for this study. However, the temperature
Fig. 11 The color intensity
indicates the temperature
distribution calculated by
SEAWAT at the end of the
injection period for cycle-2 with
density difference modification.
Black contour lines indicate the
temperature distribution
calculated by SEAWAT for
regular HT-ATES
Tabl e 7 Calculated recovery
efficiencies and relative increase
with density modification
Thermal recovery efficiency Relative recovery
increase
Salt recovery
efficiency
Regular HT-ATES Dens. diff. compensation
Cycle-1 0.58 0.68 17.2 % 0.69
Cycle-2 0.41 0.66 60.8 % 0.63
Hydrogeol J
contrasts during HT-ATES are so high that free thermal con-
vection occurs and a density equation of state is required for
the simulation. To the best of the authors’knowledge, this is
the first study that validates the approach of Langevin (2008)
for axi-symmetric density-driven groundwater flow and ther-
mal heat transport at high temperature contrasts (ΔT=40–
60 °C).
For accurate simulation of the thermal convection, a non-
linear density equation of state was implemented in the
SEAWATv4 code (Van Lopik et al. 2015). The modeling re-
sults accurately reproduce the experimental and numerical da-
ta sets of the field experiment at Auburn University (Molz
et al. 1983a;Buschecketal.1983), while using both an axi-
symmetric SEAWATv4 model domain and the incorporated
non-linear density equation of state (section ‘Modeling of the
pilot study at Auburn University’).
Salinity management of density difference compensated
HT-ATES systems
For this study, a single injection-recovery HT-ATES well
through which heat was injected and extracted was consid-
ered. Using saline water for heat storage to compensate the
density difference with the surrounding cold groundwater re-
sults in a significant increase of thermal recovery. From a
water resources and economic perspective, rather than adding
salt to the water used for HT-ATES, saline water sources could
be used, e.g. deeper, more saline groundwater, seawater, or
reverse osmosis concentrate (Pérez-Gonzaléz et al. 2012).
In practice, most HT-ATES systems use a doublet well
configuration where the hot injection-recovery well, as well
as the cold supply well are screened in the same aquifer (e.g.
Molz et al. 1983a). For such systems, the use of the density
difference method with saline water will result in salt loss due
to the salinity contrast between ambient groundwater and in-
jection water for both wells. Alternatively, the use of a mono-
well configuration (e.g. Zeghici et al. 2015) allows the cold
supply well to be screened in a deeper, more saline aquifer
which acts as the source for the required salinity.
Depending on the saline water source used for density dif-
ference compensation, the chemical composition of the water
varies. For regular HT-ATES systems, the risk of mineral-
precipitation-induced clogging of wells by, e.g. carbonate
and silica minerals is already well known (e.g. Griffioen and
Appelo 1993). For most natural saline water sources, howev-
er, salinity is mainly defined by sodium and chloride concen-
trations for which the solubility is controlled by the halite
(NaCl) mineral. Since this mineral has temperature dependent
solubilities that are well above the salt concentrations required
for the HT-ATES relevant temperature range (10–120 °C,
Fig. 2), using saline water for density difference compensation
is not expected to further increase the risk of mineral precip-
itation and clogging around the wells.
Besides the risk of temperature dependent mineral precip-
itation, a high salinity contrast between injection water and
ambient groundwater may cause clay particles in the aquifer
to swell, or shrink and migrate. Osmotic swelling of clay par-
ticles is known to occur when freshwater with a low ion con-
centration is injected into a brackish or saline aquifer (Brown
and Silvey 1977;Molzetal.1979). The use of saline water for
density difference compensation for HT-ATES in brackish or
saline aquifers as described in this study, will minimize these
effects.
Salinization of the aquifer
In this study, an ambient salinity of 10,000 ppm was assumed
for the simulated cases, since HT-ATES is currently only
allowed in brackish or saline aquifers in the Netherlands.
This is due to induced (bio)geochemical reactions during hot
water storage (Brons et al. 1991;Hartogetal.2013; Bonte
et al. 2013) that could make freshwater unusable for other
purposes. The numerical simulations in this study show that
the calculated salt mass recovery efficiencies for fully opera-
tive seasonal HT-ATES systems range from 0.59 to 0.82
(Fig 7b), resulting in a net salinization of the aquifer.
However, the risk of salinization of overlying aquifers appears
to be negligible as molecular salt diffusion and saltwater seep-
age into the upper aquitard was limited in the simulations
(Fig. 6d–f). The accumulated salt mass at the bottom of the
aquifer after thermal recovery could be recovered by a partial-
ly penetrating well screened at the bottom of the aquifer either
to minimize the salinization of the aquifer or for re-use in the
salinity management of the density difference compensation
HT-ATES system.
Summary and conclusions
Density difference compensation using saline water can be
used in HT-ATES systems to overcome the density difference
between hot injection water and colder ambient groundwater
and prevent free thermal convection of the injected hot water.
Additionally, thermal recovery efficiency is significantly in-
creased. For example, calculations for a regular seasonal HT-
ATES at a temperature of 80 °C in a high-permeability aquifer
resulted in a recovery efficiency of 0.40 for the fourth cycle,
while density difference compensation gave an efficiency of
0.69. HT-ATES with density difference compensation can be
applied in aquifers with higher hydraulic conductivities and at
larger temperatures. This means that a much broader range of
aquifers are suitable for HT-ATES and higher capacities can
be achieved.
The thermal front moves at a lower velocity than the solute
front during hot saline water injection due to thermal retarda-
tion. Consequently, downward density-driven flow of the cold
Hydrogeol J
saltwater in between the two fronts is triggered by the salt
concentration contrast between injected cold saltwater and
less saline ambient groundwater. Some local displacement of
the hot water front towards the well screen is observed, due to
the lateral transport of cold saltwater along the bottom of the
aquifer.
Saltwater accumulates at the bottom of the aquifer during the
extraction period due to continued downward density-driven
saltwater flow. Consequently, the total injected salt mass is
not fully recovered after extraction. The salt recovery efficiency
ranges from 0.59 to 0.82 for the simulated scenarios.
Axi-symmetric density-driven flow simulation in the
SEAWATv4 code with an implemented non-linear density
equation of state was validated on experimental (Molz et al.
1983a) and numerical results (Buscheck et al. 1983) of a HT-
ATES experiment conducted at Auburn University.
Acknowledgements The authors thank two anonymous reviewers for
their constructive feedback, which allowed us to improve the manuscript
significantly.
References
Bodvarsson GS (1982) Mathematical modeling of the behavior of geo-
thermal systems under exploitation. PhD Thesis, University of
California at Berkeley; and as Rep. LBL-13937, Lawrence
Berkeley Lab, Berkeley, CA
Bonte M, van Breukelen BM, Stuyfzand PJ (2013) Temperature-induced
impact on groundwater quality and arsenic mobility in anoxic aqui-
fer sediments used for both drinking water and shallow geothermal
energy production. Water Res 47(14):5088–5100. doi:10.1016/j.
watres.2013.05.049
Brons HJ, Griffioen J, Appelo CAJ, Zehnder AJB (1991)
(Bio)geochemical reactions in aquifer material from a thermal ener-
gy storage site. Water Res 25(6):729–736. doi:10.1016/0043-
1354(91)90048-U
Brown DL, Silvey WD (1977) Artificial recharge to a freshwater-
sensitive brackish-water sand aquifer. US Geol Surv Prof Pap 939
Buscheck TA, Doughty C, Tsang CF (1983) Prediction and analysis of a
field experiment on a multilayered aquifer thermal energy storage
system with strong buoyancy flow. Water Resour Res 19(5):1307–
1315. doi:10.1029/WR019i005p01307
Doughty C, Hellström G, Tsang CF, Claesson J (1982) A dimensionless
parameter approach to the thermal behavior of an aquifer thermal
energy storage system. Water Resour Res 18(3):571–587. doi:10.
1029/WR018i003p00571
Griffioen J, Appelo CAJ (1993) Nature and extent of carbonate precipi-
tation during aquifer thermal energy storage. Appl Geochem 8(2):
161–176. doi:10.1016/0883-2927(93)90032-C
Guo W, Langevin CD (2002) User’s guide to SEAWAT: a computer
program for simulation of three-dimensional variable-density
ground-water flow. US Geol Surv Tech Water-Resour Invest 6-A7
Harbaugh AW, Banta ER, Hill MC, McDonald MG (2000) MODFLOW-
2000, the US Geological Survey modular ground-water models:
user guide to modularization concepts and the ground-water flow
process. US Geol Surv Open-File Rep 00-92
Hartog N, Drijver B, Dinkla I, Bonte M (2013) Field assessment on the
impact of Aquifer Thermal Energy Storage (ATES) systems on
chemical and microbial groundwater compositions. European
Geothermal Conference, Pisa, Italy, 3–7 June 2013
Hellström G, Tsang CF, Claesson J (1979) Heat storage in aquifers: buoy-
ancy flow and thermal stratification problems. Institute of
Technology, Lund, Sweden; and as Rep. LBL-14246, Lawrence
Berkeley Lab, Berkeley, CA
Isdale JD, Morris R (1972) Physical properties of sea water solutions:
density. Desalination 10(4):329–339. doi:10.1016/S0011-9164(00)
80003-X
Kabus F, Seibt P (2000) Aquifer Thermal Energy Storage for the Berlin
Reichstagbuilding: new seat of the German Parliament. Proceedings
of the World Geothermal Congress 2000, Kyushu, Tohoku, Japan,
May 28–June 10 2000
Langevin CD (2008) Modeling axisymmetric flow and transport. Ground
Water 46(4):579–590. doi:10.1111/j.1745-6584.2008.00445.x
Langevin CD, Thorne DT Jr., Dausman AM, Sukop MC, Guo W (2008)
SEAWAT Version 4: a computer program for simulation of multi-
species solute and heat transport. US Geological Surv Tech Meth 6-
A22
Mathey B (1977) Development and resorption ofa thermal disturbance in
a phreatic aquifer with natural convection. J Hydrol 34(3–4):315–
333. doi:10.1016/0022-1694(77)90139-1
Meyer CF, Todd DK (1973) Conserving energy with heat storage wells.
Environ Sci Technol 7(6):512–516. doi:10.1021/es60078a009
Millero FJ, Poisson A (1981) International one-atmosphere equation of
state of seawater. Deep-Sea Res 28A(6):625–629. doi:10.1016/
0198-0149(81)90122-9
Miotliński K, Dillon PJ (2015) Relative recovery of thermal energy and
fresh water in aquifer storage and recovery systems. Ground Water
53(6):877–884. doi:10.1111/gwat.12286
Miotliński K, Dillon PJ, Pavelic P, Barry K, Kremer S (2014) Recovery of
injected freshwater from a brackish aquifer with a multiwell system.
Ground Water 52(4):495–502. doi:10.1111/gwat.12089
Molz FJ, Melville JG, Parr AD, King DA, Hopf MT (1983a) Aquifer
thermal energy storage: a well doublet experiment at increased tem-
peratures. Water Resour Res 19(1):149–160. doi:10.1029/
WR019i001p00149
Molz FJ, Melville JG, Güven O, Parr AD (1983b) Aquifer thermal energy
storage: an attempt to counter free thermal convection. Water Resour
Res 19(4):922–930. doi:10.1029/WR019i004p00922
Molz FJ, Parr AD, Andersen PF, Lucido VD, Warman JC (1979) Thermal
energy storage in a confined aquifer: experimental results. Water
Resour Res 15(6):1509–1514. doi:10.1029/WR015i006p01509
Oldenburg CM, Preuss K (1999) Plume separation by transient thermo-
haline convection in porous media. Geophys Res Lett 26(19):2997–
3000. doi:10.1029/1999GL002360
Olsthoorn TN (1982) The clogging of recharge wells, main subjects.
KIWA Communications-72, The Netherlands Testing and
Research Institute, Rijswijk, The Netherlands
Palmer CD, Blowes DW, Frind EO, Molson JW (1992) Thermal energy
storage in an unconfined aquifer, 1: field injection experiment. Water
Resour Res 28(10):2845–2856. doi:10.1029/92WR01471
Pérez-Gonzaléz A, Urtiaga AM, Ibáñez R, Ortiz I (2012) State of the art
and review on the treatment technologies of water reverse osmosis
concentrates. Water Res 46(2):267–283. doi:10.1016/j.watres.2011.
10.046
Réveillère A, Hamm V, Lesueur H, Cordier E, Goblet P (2013)
Geothermal contribution to the energy mix of a heating network
when using Aquifer Thermal Energy Storage: modeling and appli-
cation to the Paris basin. Geothermics 47:69–79. doi:10.1016/j.
geothermics.2013.02.005
Sanner B, Karytsas C, Mendrinos D, Ryback L (2003) Current status of
ground source heat pumps and underground thermal energy storage
in Europe. Geothermics 32(4–6):579–588. doi:10.1016/S0375-
6505(03)00060-9
Hydrogeol J
Sanner B, Kabus F, Seibt P, Bartels J (2005) Underground thermal energy
storage for the German Parliament in Berlin: system concept and
operational experiences. Proceedings of the World Geothermal
Congress, Antalya, Turkey, 24–29 April 2005
Sauty JP, Gringarten AC, Menjoz A, Landel PA (1982) Sensible energy
storage in aquifers, 1: theoretical study. Water Resour Res 18(2):
245–252. doi:10.1029/WR018i002p00245
Schout G, Drijver B, Gutierrez-Neri M, Schotting R (2014) Analysis of
recovery efficiency in high-temperature aquifer thermal energy stor-
age: a Rayleigh-based method. Hydrogeol J 22(1):281–291. doi:10.
1007/s10040-013-1050-8
Seibert S, Prommer H, Siade A, Harris B, Trefry M, Martin M (2014)
Heat and mass transport during a groundwater replenishment trial in
a highly heterogeneous aquifer. Water Resour Res 50(12):9463–
9483. doi:10.1002/2013WR015219
Sharqawy MH, Lienhard VJH, Zubair SM (2010) Thermophysical prop-
erties of seawater: a review of existing correlations and data.
Desalination Water Treat 16(1–3):354–380. doi:10.5004/dwt.2010.
1079
Vandenbohede A, Louwyck A, Vlamynck N (2014) SEAWAT-based sim-
ulation of axisymmetric heat transport. Ground Water 52(6):908–
915. doi:10.1111/gwat.12137
Van Lopik JH, Hartog N, Zaadnoordijk WJ, Cirkel DG, Raoof A (2015)
Salinization in a stratified aquifer induced by heat transfer from well
casings. Adv Water Resour 86A:32–45. doi:10.1016/j.advwatres.
2015.09.025
Voss CI (1984) A finite-element simulation model for saturated-unsatu-
rated, fluid-density-dependent ground-water flow with energy trans-
port or chemically-reactive single-species solute transport. US Geol
Surv Water Resour Invest Rep 84-4369
Wallis I, Prommer H, Post V, Vandenbohede A, Simmons CT (2013)
Simulating MODFLOW-based reactive transport under radially
symmetric flow conditions. Ground Water 51(3):398–413. doi:10.
1111/j.1745-6584.2012.00978.x
Ward JD, Simmons CT, Dillon PJ (2007) A theoretical analysis of mixed
convection in aquifer storage and recovery: how important are den-
sity effects? J Hydrol 343(3–4):169–186. doi:10.1016/j.jhydrol.
2007.06.011
Zeghici RM, Oude Essink GHP, Hartog N, Sommer W (2015) Integrated
assessment of variable density-viscosity flow for a high temperature
aquifer thermal energy storage (HT-ATES) system in a deep geo-
thermal reservoir. Geothermics 55:58–68. doi:10.1016/j.
geothermics.2014.12.006
Zheng C, Wang PP (1999) MT3DMS: a modular three-dimensional mul-
tispecies transport model for simulation ofadvection, dispersion and
chemical reactions of contaminant in ground-water systems: docu-
mentation and user’s guide. US Army Corps of Engineer Research
and Development Center, Contract Report SERDP-99-1, University
of Alabama, Vicksburg, MI
Zuurbier KG, Zaadnoordijk WJ, Stuyfzand PJ (2014) How multiple par-
tially penetrating wells improve freshwater recovery of coastal aqui-
fer storage and recovery (ASR) systems: a field and modeling study.
J Hydrol 509:430–441. doi:10.1016/j.jhydrol.2013.11.057
Hydrogeol J
