Dickinson, J., Buik, N., Matthews, M. & Snijders, A. (2008). Géotechnique 58, No. 00, 1–12 [doi: 10.1680/geot.2008.58.00.1]
Aquifer thermal energy storage: theoretical and operational analysis
FS
J. D I C K I N S O N * , N. B U I K † , M . M AT T H E W S ‡ a n d A . S N I J D E R S †
Les systèmes de stockage de chaleur en aquifère [Aquifer
thermal energy storage (ATES)] constituent une méthode
permettant d’optimiser le rendement des systèmes les
plus répandus de chauffage et de refroidissement monodirectionnels dans la nappe phréatique. Au lieu de faire
usage de la température dominante de l’eau de la nappe
phréatique soustraite, les systèmes ATES sont bidirectionnels, ce qui signifie qu’ils permettent un stockage inter saisonnier de l’énergie à haute et basse température. La
présente communication constitue une base théorique
pour l’ATES, ainsi qu’un examen empirique des performances d’un système typique installé dans un bâtiment
administratif aux Pays-Bas. Les travaux de recherche ont
été effectués par des sociétés d’ingénieur-conseil au Royaume-Uni et aux Pays-Bas, ainsi que par une université
britannique. Les conditions géologiques et hydrogéologiques examinées peuvent être décrites de façon succincte
comme un aquifère captif à sable moyen saturé, couvert
par des horizons d’argile et de sable fin. On effectue la
comparaison entre une simulation réalisée avec HSTWin,
une version modifiée du progiciel HST3D, avec des données opérationnelles recueillies au cours d’une période de
12 mois. Ces données ont été recueillies à la suite de 5
années d’exploitation. La principale conclusion de l’étude
de cas est la présence d’une bonne concordance entre la
simulation HSTWin et les relevés effectués lors de l’exploitation. En outre, on peut en déduire que des stratégies actives pour le stockage de chaleur du sol peuvent
présenter des améliorations par rapport aux systèmes
traditionnels d’énergie dont la source est dans le sol.
OO
Aquifer thermal energy storage (ATES) systems provide
a method of improving the performance of more commonly installed mono-direction groundwater heating and
cooling systems. Rather than using the prevailing temperature of the abstracted groundwater, ATES systems
are bidirectional, therefore allowing for the interseasonal
storage of low- and higher-temperature energy. This
paper provides a theoretical base for ATES and an
empirical review of the performance of a typical system
installed in an office building in the Netherlands. This
research was carried out by engineering consultancies in
the UK and the Netherlands, and a UK University. The
geology and hydrogeological conditions under focus can
be briefly described as a confined saturated medium sand
aquifer covered by horizons of clay and fine sand. A
design simulation using HSTWin, a modified version of
the software package HST3D, is compared against operational data collected over a 12-month period. These data
were collected following 5 years of operation. The main
conclusion from the case study is that there is good
agreement between the HSTWin simulation and operational findings. Furthermore, it can be inferred that
active ground thermal storage strategies can offer improvements OVER conventional ground source energy
systems.
KEYWORDS: groundwater; temperature effects; numerical
modelling and analysis; design; monitoring; environmental
engineering; sands
INTRODUCTION
The use of the ground to aid the heating and cooling of
buildings is not a new technique. However, the recent drive
towards a low-carbon economy and higher energy prices has
increased attention on the most efficient and sustainable
methods to utilise and accurately simulate these systems In
parallel, changes in building legislation have led to a move
away from traditional forms of heating and cooling commercial buildings. For example, in the UK, Building Regulations
(ODPM, 2006) have been revised in line with the European
Performance in Buildings Directive (EPBD, 2002).
Aquifer thermal energy storage (ATES) is an approach
used to enhance the efficiency in comparison with other
ground energy systems. ATES installations actively store
cooled and heated groundwater in the ground from respective heating and cooling mode cycles. For reference purposes, Fig. 1 illustrates the various systems more commonly
PR
used in practice, and highlights the position of ATES under
open loop approaches.
It is useful to make an important distinction between higherenthalpy systems, which take advantage of higher-temperature
geothermal resource generated from a heat flux originating
from the deeper geology of the earth (i.e. from decaying
isotopes of uranium, potassium and thorium in the earth’s
crust; Dickson, 2003), and lower-enthalpy systems, which take
advantage of the net solar energy absorbed and stored in the
subsurface (VDI, 2000). With the latter systems, the groundwater temperature nearer the surface is inherently linked to the
above-ground temperatures throughout the year. Within the
first few metres below ground level there are seasonal fluctuations in temperatures (Brandl, 2006). With greater depth this
effect is reduced, so the prevailing temperature is the average
annual air temperature with an additional temperature gradient
relating to the thermal flux (Rawlings, 1999).
Open-loop systems abstract and discharge groundwater
either from an aquifer of suitable permeability or from a
surface water body. Energy is then typically transferred to
and from the building’s heating and cooling system via a
heat exchanger located within the main plant room. Conversely, closed-loop systems exchange energy directly in the
ground using ground heat exchangers, which are commonly
installed in bespoke trenches at relatively shallow depth or
in vertical boreholes (VDI, 2001a). Other variations allow
for the transfer of energy through heat exchanger pipework
installed in foundation structures (Brandl, 2006).
Manuscript received
Discussion
* Faculty of Engineering and Physical Sciences, University of
Surrey, Guildford, UK; Buro Happold Consulting Engineers,
London, UK.
† IF Technology, Arnhem, The Netherlands.
‡ Department of Civil Engineering, University of Surrey, Guildford,
UK.
1
Article number = 9p022
DICKINSON, BUIK, MATTHEWS AND SNIJDERS
Low- to high-enthalpy
geothermal:
e.g. district heating,
electricity generation
Ground source
energy systems
Stored/ recharged
energy from the Sun
and
small geothermal flux
from the centre of the
Earth
Open loop:
surface water and
groundwater
abstraction/
discharge
Aquifer
bidirectional
aquifer thermal
energy storage
(ATES)
Monodirectional
Surface water
(sea/lake/river)
Closed loop:
ground loop heat
exchangers (GLHEs)
Foundation
structures:
piles/slab
OO
Surface water
(sea/lake/river)
FS
2
Horizontal trenching
Parallel piping
Spiral
Vertical boreholes
Single U-tube
Multiple U-tube
Coaxial
Fig. 1. Organisation of ground energy systems
PR
In the case of aquifer open-loop systems there are two
main techniques: monodirectional and bidirectional (VDI,
2001b). Monodirectional systems simply abstract and discharge groundwater in one direction, whereas bidirectional
systems, otherwise known as ATES systems, generally consist of two or more wells, where the flow can be reversed to
actively store heated and cooled water (Fig. 2). In summer,
groundwater is abstracted from a previously designated
‘cold’ well or wells, and is typically passed through a heat
exchanger, whereby it is possible to reject heat from the
building. The heated water is then discharged to one or more
‘warm’ wells. In winter the process is reversed, and hence it
is possible to establish a thermodynamic advantage by storing heated groundwater from the cooling summer operation
and cooled groundwater from the winter or heating operation. The thermodynamic advantage is particularly prevalent
in the cooling operation, where undisturbed groundwater
temperatures are generally not suited to direct cooling
systems in buildings that require flow temperatures of less
than 108C (CIBSE, 2004). ATES systems can eliminate the
need to run conventional cooling plant, thereby significantly
reducing running costs and the effective carbon emissions
Building
Cold well
Heat pump
Cold
thermal
plume
Heating or
cooling
Heating or
cooling
Heating or
cooling
Plate heat
exchanger
Warm well
Plant
room
Confined
aquifer
Fig. 2. Aquifer thermal energy storage system
Warm
thermal
plume
from the operation of the building. In the heating mode, heat
pumps are still typically used to upgrade the energy gained
from the ground, although the performance is now improved.
The common application of ATES remains in the lowerenthalpy field although there is an ongoing potential to
expand the theory to higher-temperature and -enthalpy systems (Sanner, 1999).
The particular problems associated with linking the
ground to the building are inherent in the geological and
hydrogeological properties of the ground beneath the site,
and the dynamic and cyclic nature of the energy loads. First,
there is a need to analyse the instantaneous heating and
cooling capacity (kW) and the cumulative required energy
transfer over a 12 month cycle (kW h). It is important at this
stage, also, to consider how the capacity and cyclic energy
transfer may change over the lifetime of the building operation. Second, owing to the variable nature of geological
thermal attributes, simulation techniques are required that
allow the subsurface energy fluxes to be modelled according
to site-specific ground parameters. Third, it is important that
a multidiscipline approach links together the respective
thermal attributes, to ensure that the system is both sustainable for the building operation and generates the desired
optimisation of the ground resource.
What is common in all low-enthalpy systems is that there
is a clear thermodynamic advantage in using residual solar
heat stored in the ground to aid heating and cooling systems
in buildings, provided the system is designed satisfactorily to
ensure long-term performance (Vinsome, 1980; Brandl,
2006; Banks, 2008). Analytical techniques to model the
ground and size closed-loop systems are well developed
(Eskilon, 1987; Hellstrom, 1991). These techniques cannot
be directly applied to open-loop systems owing to the more
complex and dynamic ground model.
This paper first reviews the theoretical basis for aquifer
thermal storage, and also the numerical solution preferred in
this case. A case study is then presented of an office
building located in the south-west of the Netherlands, which
has a two-well ATES system that has been operational for 5
years. More recently, the office building has been extensively
AQUIFER THERMAL ENERGY STORAGE
CASE STUDY: OFFICE IN DORDRECHT, THE
NETHERLANDS
Building and ATES description
The office building is located in the south-west of The
Netherlands, south of the town Dordrecht. The site plan for
the building is shown in Fig. 3.
The building is fronted with a three-storey office block,
with some packaging and manufacturing capability in a
warehouse to the rear of the building. The total treated floor
area is 4800 m2 , and the entire heating and cooling load is
delivered from the ATES system. A cold well is located in
the south-west of the site, and a warm well to the north-east
at a distance of 119 m from the cold well, as shown in
Fig. 3.
The schematic for the ATES system is shown in Fig. 4. A
primary plate heat exchanger isolates the groundwater from
the secondary side heating and cooling circuits. During
PR
As the heat rejection and abstraction is a dynamic process
throughout the year, the problem suits a step response software package with a suitable time-step resolution for the
analysis required. This must take into account the existing
condition, the step energy flux, and the natural flow through
the aquifer. The model is different from normal contaminated groundwater models owing to the thermal capacitance
of the solid matter. As a result of energy storage in this
solid fraction the thermal velocity is slower than the natural
aquifer groundwater velocity. The thermal front velocity and
time are given by (Schaetzle et al., 1980)
rw cw n
vw
(2)
vth ¼
rw cw n þ rs cs ð1 nÞ
"
#
rw cw n þ rs cs ð1 nÞ
(3)
tth ¼
tw
rw cw n
Wiedrechtse
Dike
Heat pump and
control plant
N
Office
space
Heat exchangers and
buffer tanks
Cold
well
0 10 m 20m
Fig. 3. Site plan
Space
heating HP
circuit
On reversal of the flow, the same theory can be adopted:
that is, the thermal flow is equal to the injection flow
without a variation in temperature. In practice it is accepted
that some stratification may occur in the aquifer, and heat
will be transferred by conductive processes to the confining
impermeable layers above and below, and horizontally.
HSTWin software model
To calculate the thermal impact of ATES systems the
computer model HSTWin can be used. This program has
been derived from HST3D (Heat and Solute Transport), a
program that was initially developed by the United States
Geological Survey (Kipp, 1986). Since its conception the
Warm
well
Bunsenstraat
OO
BACKGROUND TO AQUIFER THERMAL ENERGY
STORAGE
Fundamental ATES theory
Sanner (1999) reported that the first ATES modelling
research was conducted by Kazman and Rabimov in the
early 1970s (Kazmann, 1971; Rabbimov et al., 1971). However, there are records of operational ATES systems in
Shanghai, China, in the 1960s (Luxiang & Manfang, 1984).
Perhaps the first condensed publication was by Schaetzle et
al. (1980), where the basic static and time step models were
presented alongside possible applications and research activities. Since this time many systems have been successfully
installed in North America and Europe.
Thermal energy is stored in both the groundwater and
aquifer material, and hence the volumetric heat capacity is a
function of the porosity and thermal properties of the respective fluid and solid materials.
The thermal capacity of an aquifer can be defined as
h
i
(1)
Q ¼ ðrs cs Þð1 nÞ þ ðrw cw Þ n
HST3D model has been modified a number of times to
improve applicability of the software interface to ATES
system design. The program can be used not only for
simulating thermal transport through aquifers, but also for
models with differing groundwater properties, such as fresh
and salt water density. The simulation model is first validated against analytical solutions before comparing against
measured data in the case study.
FS
monitored to allow for optimisation, and also to provide data
for ongoing reporting to the Netherlands Environmental
Assessment Agency (MNP). Permits issued by the MNP for
ATES systems in the Netherlands generally require that there
be a near energy balance of heat abstracted and discharged
over a 12-month period. Twelve months of data have been
collated from 2007 to enable validation against the simulation model. This model has been updated from the initial
model, completed before project installation, to include the
2007 starting condition in the wells and the observed heating
and cooling loads in the building during this year. A
modified version of the software package HST3D has been
used to simulate the groundwater flow and thermal plume
migration at each well (Kipp, 1986).
3
v
v
Cooling
circuit
Heat
exchanger
v
Abstraction:
cooling
mode
Ground
level
Discharge:
heating
mode
Cold well
Fig. 4. ATES system schematic
Abstraction:
heating
mode
Discharge:
cooling
mode
Warm well
DICKINSON, BUIK, MATTHEWS AND SNIJDERS
heating operation water is abstracted from the warm well at
0.02 deg C/m increase in the undisturbed temperature with
approximately 148C and discharged to the cold well via the
increasing depth.
heat exchanger at approximately 68C. In the cooling season
(4)
Tw ¼ Tm þ 0:02s
the groundwater is abstracted from the cold well at approximately 7–98C and discharged to the warm well again via the
By considering a mean air temperature of 10.58C at the
heat exchanger at 13–148C. A submersible pump rated at
location, and a depth of 107.5 m at the midpoint of the well
7 kW is installed in each well, and the maximum pump flow
screen, the undisturbed temperature is calculated to be
rate is 21 m3 /h. A heat pump with a rated thermal capacity
12.78C at this site.
of 245 kW is used to generate elevated hot water at 458C,
whereas in cooling mode the prevailing flow temperatures
on the secondary side of the heat exchanger can be used
Model validation
directly for cooling in the building.
Advection is the primary method of heat transport through
the aquifer. HSTWin has been compared with several analytical solutions for heat transport through intergranular aquifers (Gringarten & Sauty, 1975; Vinsome & Westerveld,
Site geology, hydrogeology and well construction
1980). One of the analytical solutions that is used for the
The geology of the site comprises an alternating sequence
validation has been derived by Lauwerier (1955). In this
of sands and clays of Holocene and Plio-Pleistocene age,
analysis the flow is considered to be Darcian. The solution
creating a multiple aquifer system. The Holocene sediments
describes the non-stationary heat transport through an aquicomprise fluvial, tidal-flat and estuarine deposits, accumufer that is situated between two impermeable (clay) layers.
lated under the influence of a rising sea level behind a series
Only advective heat transfer is taken into account in the
of coastal barriers. These deposits reach a maximum thickaquifer, and heat conduction through the confining layers.
ness of about 25 m. In the Dordrecht area the Holocene
Fig. 7 illustrates the applied boundary conditions and system
sediments are dominated by river deposits, principally from
assumptions. If HSTWin can approximate the calculated
the Rhine. The Plio-Pleistocene sediments are dominated by
values of the analytical solution, then it follows that it can
medium to coarse fluvial sands with a thickness of about
also simulate the thermal impact of an ATES system.
150 m at Dordrecht (Breeuwer, 1973). The average hydraulic
The input parameters to compare the analytical solution
conductivity for the aquifer sands ranges from 20 to 50 m/
with the HSTWin model are given in Table 1.
day (2.3 3 104 to 5.8 3 104 m/s) (de Vries, 2007). The
Equations (5) and (6) have been derived by Lauwerier
clay layers found in both the Holocene and the Plio-Pleisto(1955)
cene sediments are not continuous, owing to the changes in
"
#
the river system over this period.
vth =
The borehole logs for the cold and warm wells at the site
(5)
T ðvth , tÞ ¼ erfc pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 ð t vth Þ=
are shown in Fig. 5. There is a reasonable correlation of
strata between each well. However, not all the beds are
ra ca 2 ð ba =2Þ2
continuous over the distance between the wells: this is
¼
(6)
typical of fluvial deposits such as these. Two distinct aquir1 cl ºl =r1 cl
fers dominated by medium-grained sand can be identified,
Figure 8 presents the comparison between the results of the
between approximately 5 and 15 mbgl, and between approxianalytical solution and HSTWin. The graph shows the ratio
mately 90 and 115 mbgl. In this case the lower aquifer
between the injection temperature and the normalised temhorizon was considered the most suitable for the application.
perature at increasing distance through the aquifer. This is
The mean nominal particle sizes (d50 ) over the length of the
assuming planar flow through the aquifer after a 20-year
well screens in the cold and warm wells are 339 m and
period.
328 m respectively (see Fig. 5).
In the HSTWin model the thermal dispersion is approxiThe groundwater flow pattern in the aquifer system is
mated to be 1/10th of the radius of the thermal store, in line
driven by rainfall-induced and topographically controlled
with field observations by Sauty et al. (1982). The thermal
potential energy (De Vries, 2007). The area around the site
dispersion correction is to account for hydrodynamic disperis relatively flat, between 0 and 5 m above sea level: hence
sion caused by unknown heterogeneities in the aquifer.
the groundwater flow rates are likely to be very low. This
A good correlation is shown through the aquifer to a
means that the warm and cold water plumes developed
distance of 200 m, where the temperature equates to the
around the wells will not be drawn out or significantly
natural undisturbed temperature in the aquifer. The agreediluted by water at the undisturbed aquifer temperature.
ment shown between the analytical solution and the HSTWin
The lowland areas of the Netherlands, such as that in
provides a degree of confidence in the software model’s
which the site is located, suffer from problems with brackish
ability to simulate heat transport in the aquifer. This is prior
and saline water. However, the water abstracted from the
to comparing the software simulation with the operational
warm and cold wells has a chloride content of less than
data from the case study.
20 mg/l, indicating the presence of fresh groundwater.
The well construction is shown in Fig. 6. The submersible
pump is located in a pump chamber constructed of PVC to
Case study modelling methodology and results
a depth of 18 m, and a smaller-diameter casing is installed
The comparison of the HSTWin with the case study was
to the screen at 100 m depth. The well screen is perforated
completed using the observed starting temperatures for 2007.
with 0.6 mm diameter slots, and is situated between 100 and
The hydraulic conductivity was set to be 20 m/day using the
115 m depth with the bottom of the well at 116 m below
equation for channel and immature sediments suggested by
ground level. A separate piezometric level sensor is installed
Sheppard (1989),
adjacent to each well. Bentonite grout is used to prevent the
:
creation of pathways for brackish water, and to isolate other
(7)
K ¼ 3500d 15065
aquifer horizons throughout the depth of the borehole.
The undisturbed or far-field groundwater temperature can
Cold well average d50 ¼ 328 m from Fig. 5: therefore
be estimated using equation (4) suggested by Eggen (1990),
K ¼ 2.6 3 104 m/s or 22.8 m/day. Warm well average d50 ¼
and presented by Rawlings (1999). This equation suggests a
PR
OO
FS
4
AQUIFER THERMAL ENERGY STORAGE
⫺10
⫺20
⫺20
Cold
FS
⫺10
5
600
0
400
0
0
200
d50: µm
600
400
200
0
d50: µm
Warm
Peat
Clay
Fine sand
⫺30
⫺30
Medium
sand
⫺60
⫺70
⫺80
Atterberg
scale
⫺50
⫺60
⫺70
⫺80
⫺90
PR
⫺90
Depth below ground level: mbgl
⫺50
⫺40
OO
Depth below ground level: mbgl
⫺40
⫺100
⫺100
⫺110
⫺110
⫺120
⫺120
Average from 100–116 ⫽ 328 µm
Average from 100–116 ⫽ 339 µm
Fig. 5. Borehole logs for cold and warm wells at site
339 m from Fig. 5: therefore K ¼ 2.8 3 104 m/s or 24.1 m/
day. The conductivity of 20 m/day used in the model is at the
lower end of the published range for this aquifer (de Vries,
2007) in order to provide a lower-bound prediction.
Figure 9 shows the respective computed monthly well
temperatures during heating and cooling modes. During the
heating season the abstraction temperature is between 13.58C
and 148C. During the period from January to April the
DICKINSON, BUIK, MATTHEWS AND SNIJDERS
6
0
Bore & 450 mm
⫺30
Sand
0
Grout
⫺50
Gravel
(2–5 mm)
⫺60
Casing
& 160 ⫻ 147·6 mm
⫺70
Filter sand
(0·8–1·25 mm)
⫺80
Piezometer standpipe
& 32/28 mm
Filter casing
⫺100
& 160 ⫻ 147·6 mm
(Slots 0·6 mm)
⫺110
Sand catcher
& 160 ⫻ 147·6 mm
⫺120
Fig. 6. Well construction
Casagrande
piezometer
60
80
100 120 140
Distance through aquifer: m
Aquifer:
heat transport predominately
through advection
Fixed
temperature
boundary
condition
Impermeable layer:
infinite thickness, heat
transport only by conduction
Fig. 7. Geological profile used for analytical Lauwerier solution:
planar flow
PR
temperature drops slightly as the thermal store is depleted.
During the cooling season, from March to November, the
abstraction temperature is calculated to be 6.58C, rising to
88C by October. The heating and cooling seasons overlap,
180
200
Heating mode: warm well abstraction
Cooling mode: cold well abstraction
Undisturbed groundwater temperature
Heating mode: cold well discharge
Cooling mode: warm well discharge
15
14
13
12
11
10
9
8
7
6
5
Jan Feb Mar Apr May Jun Jul
Months
Aug Sep Oct Nov Dec
leading to a partial recharge of the cold well from heating
operations during October and November.
Figure 10 and 11 present the data in plan view for the
start of the heating and cooling seasons respectively. The
shape is to some extent a function of the grid resolution
used in the model. In Fig. 10 the cold well has been
partially depleted, with the temperature at the well screen
approaching 88C, as also indicated by Fig. 9. At the start of
the cooling season, as shown in Fig. 11, the isothermal plot
shows the abstraction temperature to be near 6.58C. There is
a more pronounced thermal gradient from the cold well, as
the difference between the far-field temperature and the
injection temperature during the heating mode is greater.
Both isothermal contour plots suggest that there is no
thermal short-circuiting during the year. The far-field tem-
Table 1. Input parameters for HSTWin in analytical model validation
Parameter
160
Fig. 9. Computed monthly warm and cold abstraction and
discharge temperatures from HSTWin
Impermeable layer:
infinite thickness, heat
transport only by conduction
Fixed
temperature
boundary
condition
40
OO
⫺90
20
Fig. 8. Temperature profile through confined aquifer: analytical
solution for heat transport through two confining layers
compared with predicted values from HSTWin
Temperature: °C
⫺40
Analytical
FS
T/Tinj
⫺10
Pump chamber
PVC & 250 ⫻ 230·8 mm
⫺20
HSTWin
1·0
0·9
0·8
0·7
0·6
0·5
0·4
0·3
0·2
0·1
0
Value
Length of HSTWin model, x: m
800
Temperature difference between injection and initial (aquifer) temperature, ˜T: K
4
Node distance; ˜x: m
1
Volumetric heat capacity, solid material: rs cs : MJ/(m3 K)
1.5
Volumetric heat capacity, (fresh) water: rw cw : MJ/(m3 K)
4.2
Volumetric heat capacity, over-/underlying layer (water saturated), r1 cl : MJ/(m3 K)
2.5
Heat conduction, aquifer, ºa : W/(m K)
2.5
Heat conduction, over-/underlying layer: ºl : W/(m K)
1.7
Porosity, n
0.35
Thickness of aquifer, ba : m
20
Hydraulic gradient, K: m/day
20
Gradient, i
1
Maximum time step, ˜t: h
0.5
AQUIFER THERMAL ENERGY STORAGE
100
tion temperature from the cooling season but also below the
far-field temperature of 12.78C. This would suggest that the
thermal plume from the cold well had migrated to the warm
well.
Warm well
60
North–south: m
40
13
.5
13
°C
20
0
⫺20
8
⫺40
⫺60
Cold well
9
10 1
1
°C
12
⫺80
⫺100
⫺100 ⫺80 ⫺60 ⫺40 ⫺20
0
20
West–east: m
40
60
80
100
100
80
Warm well
Q_ ¼ m_ w cw ˜T
60
North–south: m
40
13
.5
20
13
°C
0
⫺20
⫺40
⫺60
⫺80
Operational data summary
Various measurements were taken in the building at 8 min
intervals throughout the year and recorded via a building
management system software package. The data were then
imported into a spreadsheet to allow further analysis and
interpretation. A summary of the data points used in the
analysis is shown in Table 2.
The prevailing ambient air temperatures for the site were
measured via a temperature sensor externally located to the
north of the building to avoid adverse measurements due to
direct sunlight. The average monthly daily recorded temperatures are presented in Fig. 12, alongside the average minimum and maximum daily temperatures. During 2007 the
maximum recorded absolute temperature was 41.18C and the
minimum was 5.68C.
The observed energy and volumetric abstraction during
the heating mode and the energy rejection during the cooling
mode are shown in Figs 13 and 14 respectively. The energy
flows were calculated on the primary side of the heat
exchanger using an obtrusive flow meter, and temperature
probes located in the pipe flow on each side. The energy
flow is calculated using the equation
OO
Fig. 10. Computed isothermal contour plot from HSTWin:
September, start of heating season
FS
80
7
Cold well
7
8
9
101
1 C
12°
⫺100
⫺100 ⫺80 ⫺60 ⫺40 ⫺20
0
20
West–east: m
40
60
80
100
PR
Fig. 11. Computed isothermal contour plot from HSTWin:
April, start of cooling season
perature has been calculated to be 12.78C. If short-circuiting
was demonstrated, the contour plot for either well would
start to overlap the adjacent well. For example, part-way into
the heating season the abstraction temperature from the
warm well would fall significantly, not only below the injec-
(8)
It can be deduced that the heating and cooling operations
overlap during March to April and September to November.
During these mid-season periods the building occupier
manually changes the direction of flow between the two
wells according to the building demands. The cumulative
heat energy abstracted was 86.4 MW h, and the total cooling
energy was 64.4 MW h: hence there was a net heat energy
abstraction from the aquifer of 22 MW h observed during
2007. In comparison the cumulative volume of groundwater
abstracted from the cold well was 10 930 m3 compared with
9760 m3 abstracted from the warm well, a difference of
1170 m3 .
The variable-speed submersible pump abstracts groundwater from the respective well according to the relative
demands of the building. Figs 15 and 16 illustrate the typical
pumping regime, flow rates and well levels during heatingmode operation on 2 January 2007. After a short time the
temperatures monitored before the primary side heat exchanger reach steady state during pumping, that is, equate to the
actual groundwater temperature. When the well pump is
turned off, the temperature tends towards the ambient air
temperature in the plant room. Unfortunately, no measurements were taken of the ambient plant room temperature.
During steady-state operation the temperature differential is
approximately 88C. The flow rate fluctuates continuously
Table 2. Summary of measurements taking during monitoring of ATES system at
Dordrecht
Measurement
Sensor location
External air temperature
Cold well abstraction volume
Warm well abstraction volume
Temperature, warm well
Temperature, cold well
Warm well level
Cold well level
External wall to north of building
Plant room: adjacent to plate heat
Plant room: adjacent to plate heat
Plant room: adjacent to plate heat
Plant room: adjacent to plate heat
Adjacent to well
Adjacent to well
exchanger
exchanger
exchanger
exchanger
DICKINSON, BUIK, MATTHEWS AND SNIJDERS
25 000
20 000
⫺1
⫹sw
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Months
Warm well abstraction
3500
Cold well abstraction
Flow rate: m3/h
2500
2000
1500
1000
500
08:00
Groundwater abstraction flow rate
Warm well discharge temperature
Cold well abstraction temperature
10
8
6
4
Groundwater flow rate
Warm well abstraction temperature
22:00
20:00
18:00
16:00
14:00
12:00
10:00
08:00
06:00
04:00
Warm well drawdown
Cold well drawdown
Time: h
Fig. 15. Measured heating mode flow rate and abstraction and
discharge temperature on 2 January 2007
⫺SW
1
0
⫺1
20:00
18:00
16:00
14:00
12:00
10:00
02:00
⫺3
08:00
⫹SW
⫺2
06:00
Drawdown: m
2
Temperature: ºC
18
16
14
12
10
8
6
4
2
0
00:00
22:00
20:00
18:00
16:00
14:00
12:00
10:00
08:00
06:00
04:00
Time: h
3
Cold well discharge temperature
02:00
20
18
16
14
12
10
8
6
4
2
0
Fig. 17. Measured cooling mode flow rate and abstraction and
discharge temperature from 13 July 2007
04:00
PR
Fig. 14. Measured monthly warm and cold well groundwater
abstraction volume
02:00
2
0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Months
00:00
06:00
Fig. 16. Measured heating mode well level differential from 2
January 2007
12
0
Flow rate: m3/h
Time: h
14
3000
20
18
16
14
12
10
8
6
4
2
0
04:00
5000
02:00
⫺3
10 000
Fig. 13. Measured monthly groundwater energy abstraction and
rejection
Groundwater volume: m3
0
⫺2
15 000
0
⫺sw
1
22:00
Cooling energy
Cold well drawdown
2
Drawdown: m
Heating energy
30 000
Warm well drawdown
3
OO
Energy abstraction/rejection: kWh
Fig. 12. Measured monthly average ambient air temperature,
2007
Temperature: ºC
Aug Sep Oct Nov Dec
22:00
Jan Feb Mar Apr May Jun Jul
Months
00:00
0
20:00
5
18:00
10
16:00
15
14:00
20
00:00
Temperature: °C
25
12:00
30
FS
throughout the day to match the operation of the heat pump,
which in turn modulates to charge an in-series thermal
buffer vessel. The maximum recorded flow from the warm
well was 19.1 m3 /h.
During pumping the respective well levels were measured.
Fig. 16 shows that the typical drawdown in the well for
pumping at a flow rate of 15 m3 /h is ,1 m from a resting
water level of 3.1 mbgl. During 2007 the lowest recorded
level in the warm well was 4.7 mbgl, corresponding to a
drawdown of 1.6m.
It can be seen from Figs 17 and 18 that in the cooling
mode the flow rate fluctuates less during the day as there is
Average daily temperature
Average maximum daily temperature
Average minimum daily temperature
10:00
8
Time: h
Fig. 18. Measured cooling mode well level differential from 13
July 2007
AQUIFER THERMAL ENERGY STORAGE
Tw,out ¼ Twall þ ð Tw,source Twall Þeð hA= mw cw Þ
PR
Nu:ºw
(10)
D
For the purposes of this investigation the mean temperature
at the pipe wall (Twall ) for the length of the well shaft has
been approximated using equation (4) and a depth of 58 m.
This reflects the midpoint of the well depth. Along the
length of the horizontal trench the temperature, Twall, can be
approximated using the equation (Carslaw & Jaeger, 1959)
h
i
Twall ð s, tÞ ¼ Tm Tamp e sð=365ÆÞ1=2
n
h
io
3 cos 2=365 t t0 s=2ð365=ÆÞ1=2
and
Temperature: °C
FS
Estimated pipe wall temperature in well
18
16
14
12
10
8
6
Jan Feb Mar Apr May Jun Jul
Months
Aug Sep Oct Nov Dec
Table 3. Parameters for source temperature correction
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Months
Fig. 19. Measured monthly average warm
abstraction temperatures
Estimated pipe wall temperature at 1·5 m
Fig. 20. Estimated pipe wall temperature along horizontal
trench at 1.5 mbgl and along length of well
Heating mode: warm well abstraction
Cooling mode: cold well abstraction
Undisturbed groundwater temperature
Heating mode: cold well discharge
Cooling mode: warm well discharge
15
14
13
12
11
10
9
8
7
6
5
(11)
Equation (11) has been used to calculate the temperature at
1.5 mbgl for a period of 12 months using the corresponding
average monthly air temperatures for 2007 shown in Fig. 12.
The calculated pipe wall temperatures for the well shaft and
at 1.5 mbgl are shown in Fig. 20.
Using these estimated pipe wall temperatures the predicted
abstraction temperatures from HSTWin have be calculated
using (9) and (10). The parameters used in this approximation are shown in Table 3. The corrected HSTWin temperatures can now be compared with the HSTWin model outputs
and the measured values.
The comparisons are shown in Figs 21 and 22. There is
now an improved correlation with the cold well abstraction
temperature and the measured data from the plant room.
There is also an improvement in the warm well abstraction
predicted temperatures, although, admittedly, from January
to April the improvement is marginal. To be more accurate
the correction would require a more dynamic ground model
to account for conductive heat abstraction from surrounding
Temperature: ºC
DISCUSSION
Comparison of computed and measured data
The predicted and measured data for the cooling and
heating modes are shown in Figs 9 and 19 respectively. In
the cooling mode the measured temperature is higher than
predicted by the model. In the heating mode the difference
is less pronounced. The temperature was measured in the
plant room, whereas the model predicts the temperature at
the well screen. The positioning of the temperature sensors
was initially chosen by the building designers to monitor the
groundwater temperature adjacent to the heat exchangers.
This is ultimately of greater interest than the well screen
temperature when implementing a control strategy for heating and cooling plant.
In the cooling mode the abstraction temperature is lower
than the bulk temperature of the ground around the well
shaft, and also along the horizontal trench between the well
head and the inlet to the building. Therefore, as the groundwater flows from the base of the well to the plant room, the
temperature is likely to increase. The pipework is only
insulated within the building.
To investigate this further, the temperature uplift can be
approximated using the equations (de Paepe, 2003)
(9)
h¼
OO
no buffer vessel installed for the passive cooling operation.
The steady-state temperature differential is approximately
68C. In this example, with a flow rate of 6–10 m3 /h, the
drawdown was recorded at between 0.4 and 0.6 m from a
resting water level at 3.4 mbgl. During 2007 the maximum
recorded flow from the cold well was 16.5 m3 /h. By way of
comparison with the warm well operation at 15 m3 /h, the
drawdown was recorded at 0.7–0.9 m, although this was
realised only for relatively short isolated periods of 8 and
16 min. The maximum drawdown observed throughout the
year in the warm well was 1.4 m, and 1.3 m in the cold
well.
Figure 19 shows the measured average abstraction and
discharge temperatures for the warm and cold wells. Towards the end of the heating season in March and April
there is a slight drop in the steady-state abstraction temperature as the groundwater thermal store is depleted. The
abstraction temperature from the warm well is nearer 148C
following the cooling season. The depletion of the cooling
energy storage is more pronounced towards the end of the
cooling operation. This is reflected by the rising abstraction
temperature from March to October. The abstraction temperature is again reduced in November following a period of
heating and cold water discharge.
9
cold well
Cold well
Warm well
Length of horizontal
pipe ¼ 65 m
V_ ¼ 8 m3 /hr
m_ ¼ 2.22 kg/s
Length of horizontal
pipe ¼ 55 m
V_ ¼ 16 m3 /hr
m_ ¼ 4.44 kg/s
cw ¼ 4.2 kJ/(kg K)
ºw ¼ 0.59 W/(m K)
Groundwater pipe inner diameter ¼ 0.15 m
Depth to midpoint of well screen ¼ 107.5 m
Nu ¼ 3.66 (laminar flow)
DICKINSON, BUIK, MATTHEWS AND SNIJDERS
Predicted
Measured
Corrected
9·5
Temperature: ºC
9·0
8·5
8·0
7·5
7·0
6·5
6·0
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Months
Fig. 21. Cold well abstraction temperatures: HSTWin (predicted) compared with measured and HSTWin (corrected)
Predicted HSTWin
Measured
Corrected HSTWin
15·5
14·0
13·5
13·0
12·5
12·0
OO
Temperature: ºC
15·0
14·5
hydraulic conductivity and groundwater chemistry. In circumstances where a significant hydraulic gradient is anticipated (i.e. from the completion of a desktop study), the
equipotential surface can be deduced using piezometers at a
minimum of three locations, as detailed, for example, by
Freeze & Cherry (1979). In the case study the hydraulic
gradient was considered to be negligible at the desktop study
stage. The measured temperatures during 2007 did not
contradict this assumption, although the incidence of a significant hydraulic gradient cannot be proved without carrying
out more detailed studies of the transport of heat at the site.
This might prove too difficult to do with continued building
occupancy and without installing further observation wells.
However, HSTWin does have the ability to use this information to assess the movement of thermal energy over distinct
time periods linked to the heating and cooling strategy of a
building.
There is also suitable literature to correlate near-homogeneous aquifer properties (Sheppard, 1989) from standard
borehole logging techniques. However, in situ testing still
offers the obvious potential to increase the accuracy of input
data for the HSTWin model. In some countries, for example
in the Netherlands and the UK, there is also an absolute
requirement to carry out such testing before gaining an
abstraction licence from the Environment Agency. For heterogeneous fractured bedrock systems the applicability of
desk study data is severely limited due to the inherent
randomness of fissured systems. In such circumstances testing data for each specific well will be paramount to calculate
the abstraction and discharge parameters required for the
HSTWin model.
FS
10
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Months
Fig. 22. Warm well abstraction temperatures: HSTWin (predicted) compared with measured and HSTWin (corrected)
sand backfill and soil in the horizontal trench and clay, and
sand and intermediate gravel backfill in the well. Also, the
PVC piping will need to be taken into account, as this
provides a thermal barrier to heat transfer. This could be
completed using theory developed for closed-loop GSHP
systems by Eskilon (1987) and Hellstrom (1991). From this
viewpoint it is likely that the correction partially exaggerates
the heat transfer between the ground and the abstracted
groundwater. However, in the absence of this dynamic model
a simpler approximation method can be used to help calibrate predicted abstraction temperatures to design and optimise the heating and cooling system more accurately.
PR
Significant limiting factors for model
The numerical solution presented is limited in its application to the simulation of saturated aquifers, either unconfined
or confined. Difficulties are likely to arise in unsaturated
aquifers owing to the inherent inconsistency in heat capacity
throughout the store. Also, higher-temperature heat storage,
possibly from waste heat from combined heat and power
plants or solar thermal systems, may reduce the accuracy of
the model and the recovery efficiency without careful consideration of the input parameters and the possibility of
chemical precipitation at higher temperatures.
The type of analysis used in this instance is limited to
intergranular aquifers where the flow is assumed to be
Darcian. Also, the total porosity is assumed to be the same
as the total porosity. A different approach is needed for
fractured bedrock systems where the flow regime cannot be
assumed to be homogeneous.
Site investigation
Site investigation techniques are available to assess the
necessary ground parameters for modelling, including the
CONCLUSIONS
The primary objectives of this paper were to outline the
fundamental theory of ATES systems and validate the application of numerical solutions to groundwater thermal energy
storage in confined homogeneous aquifers. ATES systems
can be shown to offer thermodynamic advantages over
monodirectional systems that do not actively prioritise the
seasonal storage of heat. Software can be adapted to accurately model the cyclic nature of heat abstraction and rejection from the operation of a building heating and cooling
system.
Using the measured data from the case study, reasonable
correlation was achieved between the observed operational
data and the HSTWin model. This was true when the
absolute heat abstraction and rejection input data were
known. No thermal interaction was calculated or observed,
although the cold well store was depleted towards the end of
the cooling season. Improved correlation was achieved by
modifying the predicted HSTWin abstraction temperatures
from the cold and warm wells to account for heat transfer
from the base of each well to the plant room. This was
completed using an approximation of ground temperatures
along the well length and the horizontal trench.
During the design process the accuracy of design data
needed for HSTWin, and in particular the anticipated heating
and cooling loads in the building and hence the respective
heat storage in the ground, is more likely to impact on the
long-term performance of the system. Very good correlation
has been proved in this case study using measured data, but
with new buildings estimations have to be made regarding
the future occupancy and thermal performance of the building. In this respect, a multidiscipline interaction is essential
between the various project engineers to ensure that the
sizing of the system is adequate to account for the variability between the design case and the actual operation
characteristics of the final building. Sensitivity analysis is
AQUIFER THERMAL ENERGY STORAGE
ACKNOWLEDGEMENTS
This research has been funded primarily by the Engineering and Physical Sciences Research Council in the UK, Buro
Happold Consulting Engineers and IF Technology. The
collection of data and information from the case study
building was made possible with the kind cooperation of
Bas Kuipers of HVL.
NOTATION
temperature at pipe wall
time, thermal front in aquifer
time, groundwater
time phase lag
volumetric flow
specific volumetric flow rate
velocity, thermal front in aquifer
velocity, groundwater
soil thermal diffusivity
time constant
heat conduction, impermeable layer
thermal conductivity, aquifer
thermal conductivity of groundwater
density, impermeable layer
density, aquifer
density, solid material
density, groundwater
FS
Twall
tth
tw
t0
V
V_
vth
vw
Æ
º1
ºa
ºw
r1
ra
rs
rw
REFERENCES
Banks, D. (2008). An introduction to thermogeology: Ground source
heating and cooling. Oxford: Blackwell.
Brandl, H. (2006). Energy foundations and other thermo-active
ground structures. Géotechnique 56, No. 2, 81–122.
Breeuwer, J. B. & Jelgersmer, S. (1973). An east-west geo-hydrological section across The Netherlands. Verhandelingen Koninklijk Nederlands Geologisch en Mijnbouwkundig Genootschap 29,
105–106.
Carslaw, H. S. & Jaeger, J. C. (1959). Conduction of heat in solids,
2nd edn. Oxford: Oxford University Press.
CIBSE (2004). Refrigeration: CIBSE Guide B4. London: Chartered
Institute of Building Services Engineers.
De Paepe, M. & Janssens, A. (2003). Thermo-hydraulic design of
earth-air heat exchangers. Energy and Buildings 35, No. 3,
389–397.
De Vries, J. J. (2007). Groundwater. In Geology of the Netherlands
(eds T. E. Wong, D. A. J. Batjes and J. de Jager), pp. 295–315.
Amsterdam: Royal Netherlands Academy of Arts and Sciences.
Dickson, M. H. & Fanelli, M. (2003). Geothermal energy: Utilisation and technology. Paris: Unesco Publishing.
Eggen, G. (1990). Lavtemperatur Varmekilder. Program for anvendelse av varmepumper: Temakurs for konsulenter. Oslo: Norges
Teknisk-Naturvitenskapelige Forskningsräd.
EPBD (2002). Directive on the energy performance of buildings,
2002/91/EC. Brussels: European Parliament.
Eskilon, P. (1987). Thermal analysis of heat extraction boreholes.
PhD thesis, Department of Mathematical Physics, University of
Lund, Sweden.
Freeze, R. A. & Cherry, J. A. (1979). Groundwater. Englewood
Cliffs, NJ: Prentice Hall.
Gringarten, A. C. & Sauty, J. P. (1975). A theoretical study of heat
extraction from aquifers with uniform regional flow. J. Geophys.
Res. 80, No. 35, 4956–4962.
Hellstrom, G. (1991). Ground heat storage: Thermal analyses of
duct storage systems. PhD thesis, Department of Mathematical
Physics. University of Lund, Sweden.
Kazmann, R. G. (1971). Exotic uses of aquifers. J. Irrigat. Drainage Div. ASCE 97, No. IR3, 515–522.
Kipp, K. L. (1986). HST3D: A computer code for simulation of
heat and solute transport in three-dimensional-ground-water
flow systems. US Geological Survey Water Resources Investigations Report 86-4095. Reston, VA: United States Geological
Survey.
Lauwerier H. A. (1955). The transport of heat in an oil layer caused
by the injection of hot fluid. App. Sci. Res. 5, Nos. 2–3, 145–
150.
Luxiang, S. & Manfang, B. (1984). Case history No. 9.2: Shanghai,
China. In Guidebook to studies of land subsidence due to
ground-water withdrawal (ed. J. F. Poland), pp. 155–160. Paris:
UNESCO.
ODPM (2006). Building Regulations: Conservation of fuel and
Power Part L. London: Office of the Deputy Prime Minister.
Rabbimov, R. T., Umarov, G. Y. & Zakhidov, R. A. (1971). Storage
of solar energy in sandy-gravel ground. Geliotekhnika 7, No. 5,
57–64.
OO
therefore needed to understand the limits of operation, and
simulation remains a valid way to understand how variations
from the design case will affect the performance in the short
and long term.
ATES is a valid way to improve performance compared
with monodirectional systems. With cold groundwater storage from the heating season, lower temperatures can be
realised that enable direct cooling. In the heating mode,
raised groundwater temperatures improve the performance of
the heat pump plant. There is also a benefit compared with
closed-loop systems, where underground thermal energy
storage is limited by the principal conductive process of
these systems.
Regulatory aspects are likely to lead to the stricter control
of closed- and open-loop systems in all countries where
there is a thermodynamic advantage of using such techniques, that is, mainly temperate climates where there are
significant seasonal swings in temperature. In The Netherlands the environmental authorities currently specify that all
ATES systems must be thermally balanced. Building occupiers are required by law to inform the environmental
authorities each year of the respective volumetric and energy
abstraction and rejection. This methodology is justified to
ensure the long-term performance and prevent the shortcircuiting of individual systems, and also prevent possible
impacts on nearby installations utilising the same aquifer.
For the user the long-term performance can also be ensured,
assuming the building continues to operate within certain
limits. Following the Dutch example, it is likely that many
countries will specify numerical simulation as a requirement
prior to approval by the respective governing body.
PR
A cross-sectional area of pipe from base of well to plant
room
ba aquifer thickness
c1 heat capacity, impermeable layer
ca heat capacity, aquifer
cs heat capacity, solid material
cw heat capacity, groundwater
D pipe diameter
d50 nominal particle diameter in m for 50% of particles
smaller in particle size distribution curve
erfc complementary error function
H heat convection coefficient
m_ w mass flow rate of groundwater
N porosity
Nu Nusselt number
K hydraulic conductivity
Q thermal energy
Q_ thermal energy rate
S depth below ground level
sw drawdown
T time; temperature
Tamp annual amplitude of temperature fluctuation at ground
surface
Tinj initial injection temperature
Tm mean annual temperature at ground surface
Tw mean undisturbed groundwater temperature
Tw,out corrected HSTWin groundwater temperature
Tw,source predicted HSTWin groundwater temperature
11
DICKINSON, BUIK, MATTHEWS AND SNIJDERS
VDI (2000). Thermische Nutzung des Untergrundes: Grundlagen,
Genehmigungen, Umweltaspekte (Thermal use of the underground: Fundamentals, approvals and environmental aspects),
VDI 4640 Part 1. Berlin: Beuth Verlag.
VDI (2001a) Thermische Nutzung des Untergrundes: Erdgekoppelte
Wärmepumpen (Thermal use of the underground: Ground source
heat pump systems), VDI 4640 Part 2. Berlin: Beuth Verlag.
VDI (2001b) Thermische Nutzung des Untergrundes: Unterirdische
Thermische Energiespeicher (Utilization of the subsurface for
thermal purposes: Underground thermal energy storage), VDI
4640 Part 3. Berlin: Beuth Verlag..
Vinsome, P. K. W. & Westerveld, J. (1980). A simple method for
predicting cap and base rock heat losses in thermal reservoir
simulators. J.. Can. Pet. Technol. 19, No. 3, 87–90.
PR
OO
Rawlings, R. H. D. (1999). Ground source heat pumps: A technology overview, Technical Note 18/19. Bracknell: BSRIA.
Sanner, B. (1999). High temperature underground thermal energy
storage: State of the art and prospects, Giessener Geologische
Schriften, No. 67. Giessen: Lenz-Verlag.
Sauty, J. P., Gringarte, A. C., Fabris, N. H., Thiery, D., Menjoz, A.
& Landel, P. A. (1982). Sensible energy storage in aquifers:
field experiments and comparison with theoretical results. Water
Resour. Res. 18, No. 2, 253–265.
Schaetzle, W. J., Brett, C. E., Grubbs, D. M. & Seppanen, M. S.
(1980). Thermal energy storage in aquifers: Design and applications. New York: Pergamon.
Sheppard, R. G. (1989). Correlations of permeability and grain size.
Ground Water 27, No. 5, 633–638.
FS
12