Energy 64 (2014) 747e757

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

G-Functions for multiple interacting pile heat exchangers

Fleur Loveridge*, William Powrie
Faculty of Engineering & Environment, University of Southampton, UK

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 31 May 2013
Received in revised form
31 October 2013
Accepted 6 November 2013
Available online 8 December 2013

Pile heat exchangers e where heat transfer pipes are cast into the building piled foundations e offer an
opportunity to use ground energy systems without the additional construction costs related to the
provision of special purpose heat exchangers. However, analysis methods for pile heat exchangers are
still under development. In particular there is an absence of available methods and guidance for the
amount of thermal interaction that may occur between adjacent pile heat exchangers and the corresponding reduction in available energy that this will cause. This is of particular importance as the locations of foundation piles are controlled by the structural demands of the building and cannot be
optimised with respect to the thermal analysis. This paper presents a method for deriving G-functions for
use with multiple pile heat exchangers. Example functions illustrate the primary importance of pile
spacing in controlling available energy, followed by the number of piles within any given arrangement.
Signicantly it was found that the internal thermal behaviour of a pile is not inuenced appreciably by
adjacent piles.
2013 Elsevier Ltd. All rights reserved.

Keywords:
Ground heat exchanger
Pile
Ground energy system
Ground source heat pump system

1. Introduction
Installations of ground energy systems e where ground source
heat pumps are connected to a series of ground heat exchangers to
obtain sustainable heating and cooling e are increasing at a signicant rate. Lund et al. [1] report that the energy provided by such
systems has more than doubled between 2005 and 2010. As the
market has expanded, so has the range of types of heat exchanger
used to provide the heat source. Traditionally, elds of borehole
heat exchangers have been used to provide a large thermal capacity
for major commercial or municipal buildings. However, where
buildings require deep foundations it is becoming more common to
make dual use of the piled foundations, equipping them with heat
transfer pipes so that they can act as heat exchangers as well as
structural components, e.g. Refs. [2e4]. This can lead to cost savings
on projects by removing the need to construct special purpose heat
exchangers.
Design methods for elds of borehole heat exchangers are well
developed. Numerous different commercial software packages are
available, some of which will be mentioned below. Typically these
methods are also applied for the design of pile heat exchangers.

* Corresponding author. Faculty of Engineering & Environment, University of

Southampton, Higheld, Southampton SO17 1BJ, UK. Tel.: 44 (0)2380592662.
E-mail address: [email protected] (F. Loveridge).
0360-5442/$ e see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.energy.2013.11.014

However, there are important differences between borehole and

pile heat exchangers which mean that application of borehole
methods will underestimate the energy available from piles [5].
Heat transfer problems are typically controlled by their geometry
and the thermal properties of the materials. The thermal properties
of the materials for borehole and concrete pile construction are in a
similar range, but their geometries are not. Piles have a much
smaller aspect ratio (AR (aspect ratio) heat exchanger length/
diameter) than boreholes, which affects their long term thermal
characteristics [6]. In addition, the larger diameter of piles results in
a different short term behaviour with a greater proportion of short
term energy storage happening within the heat exchanger itself,
rather than just in the ground [5].
Previous work on pile heat exchanger design has focused mainly
on determining the temperature changes with time, usually
expressed as temperature response functions or G-functions, for a
single pile [5,7,8]. However, all ground energy systems using pile
heat exchangers will comprise multiple heat exchangers which will
interact. With the exception of the Duct Storage Model [9] (which
will be discussed below) no other specic solution for interacting
pile heat exchangers has been published. This paper builds on
previous work [5] and presents G-functions for multiple pile heat
exchangers in some simple congurations, together with an
approach for determining G-functions for any arrangement of piles.
The impact on the available energy of different pile arrangements
and spacings will be discussed, along with strategies for maximising the energy output.

748

F. Loveridge, W. Powrie / Energy 64 (2014) 747e757

Rc
Rp
rb
Tb
Tf
Thp-in
Thp-out
Ti-in
Ti-out
Tp
t

Notation
AR
B
c
Gc
Gg
Fo
H
_ hp
m
n
Q
q
Sc
Scv

aspect ratio
pile spacing
concrete cover
concrete G-function
pile G-function
Fourier number or normalised time
pile length
mass ow rate of uid at the heat pump
number of piles or pipes
power
power per metre depth
specic heat capacity
volumetric heat capacity

ag
F
lc
lg

2. Background
2.1. Design of single heat exchangers
G-Functions describe the change in temperature in the ground
around a heat exchanger with time as the result of an applied
thermal load. Usually both time and temperature change are normalised. Other normalisations are used, but this paper applies the
following expressions, based on precedent from other studies, e.g.
Refs. [7,10]. The temperature change (DT) is normalised by the
applied thermal load q (in W/m) and the thermal conductivity of
the ground lg (in W/m K) such that:

2plg
DT
q

(1)

The elapsed time (t) is normalised by the ground thermal

diffusivity (ag) and the heat exchanger radius (rb):

Fo

ag t
rb2

(2)

Eskilson [11] pioneered the G-function concept. He assumed

that ground heat exchangers act like a nite line heat source,
operating within a uniform medium with an initial temperature
equal to the far eld boundary conditions and also the surface
boundary condition. As a result of the constant temperature surface
boundary condition the temperature response (F) will reach steady
state at large values of time as the heat output is matched by that
available. The time at which this occurs and the equilibrium temperature change depends on the aspect ratio of the heat exchanger.
Eskilsons G-functions, published for AR 1000, were developed by
means of a combination of analytical and numerical methods and
form the basis of many commercial software packages.
Another commonly adopted approach is the cylindrical heat
source [12,13]. This is identical to the line heat source in the longer
term, but accounts for the nite radius of the heat exchanger in the
short term. This approach has been applied to the design of pile
heat exchangers due to their larger diameter, but it has recently
been shown [14] that the either the line heat source approach or
the solid cylinder approach is more appropriate. While the classic
cylindrical heat source assumes that the heat source is hollow, the
solid cylinder model [8] assumes that heat may also ow into the
centre of the heat exchanger. This makes it much more suitable for
simulating pile heat exchangers.

pile concrete thermal resistance

pipe thermal resistance
heat exchanger radius
temperature at pileeground interface
mean uid temperature
temperature of uid entering heat pump
temperature of uid leaving heat pump
temperature of uid entering an individual pile
temperature of uid leaving an individual pile
temperature at pipes
elapsed time
thermal diffusivity of ground
normalised temperature change
thermal conductivity of concrete
thermal conductivity of ground

All the above approaches take no account of the arrangement of

heat transfer pipes within the pile cross-section. Only two approaches overcome this limitation. Li & Lai [7] published G-functions based on the superposition of multiple line heat sources, with
each line source representing a single pipe. This can produce
theoretically accurate G-functions, but has the disadvantage of
requiring a complicated calculation process for each specic pile
diameter and pipe arrangement. This will be a barrier to its adoption on a routine basis. Loveridge & Powrie [5] presented upper and
lower bound pile G-functions based on numerical analysis of a
range of commonly constructed pile geometries (Fig. 1). This
removes the need to make new analytical solutions for every geometry but there is a trade-off in terms of a small reduction in
accuracy.
Traditionally G-functions, describing the temperature change
within the ground, are combined with a steady state resistance
which determines the temperature change between the heat
transfer uid and the ground. However, recent work [15] has shown
that this approach is conservative for pile heat exchangers and that
greater energy efciency can be obtained by taking a transient
approach to the temperature change within the heat exchanger.
Consequently Loveridge & Powrie [5] also proposed concrete Gfunctions which described the proportion of the steady state pile
resistance which is appropriate for different values of normalised
time, Fo.
2.2. Multiple heat exchangers
If ground heat exchangers are positioned within a certain
proximity of each other they will interact thermally to an increasing
degree as time passes. If not accounted for in design, this will
reduce the sustainability of a ground energy system over time. All
multiple heat exchanger G-functions are based on the principle of
spatial superposition. Eskilson [11] developed G-functions for many
different arrangements of borehole heat exchangers. Software
packages, such as EED (Earth Energy Designer) [16] and GLHEPRO
[17] are based on these, and typically use interpolation to cater for
borehole arrangements different from those given in Ref. [11].
However, in all cases Eskilsons published G-functions are limited to
the case where AR 1000 and are therefore not suitable for use
with pile heat exchangers which typically have AR < 50 [6]. In
addition, Hellstrm et al. [16] have shown EED to be unreliable for
short heat exchangers (<15 m); this effect has also been observed
by Wood et al. [18,19]. The proprietary software Orpheus builds on
GLHEPRO and overcomes some of the limitations by carrying out

F. Loveridge, W. Powrie / Energy 64 (2014) 747e757

749

Fig. 1. Upper bound (left) and lower bound (right) pile geometries and material properties, assumes pipe diameter of 25 mm.

direct superposition for each specic arrangement of heat exchangers as required, removing the need to interpolate between
published G-functions. It also claims to be more applicable to piles
of larger diameter, but further details are not given [20]. Similar
superposition approaches can be taken with cylindrical source
models, e.g. Refs. [21,22] or equally any other single heat exchanger
G-functions.
An alternative approach to G-functions for multiple heat exchangers is the so called Duct Storage Model or DST [9,23,24]. This
assumes that a large number of vertical heat exchangers are
installed close together to act as an underground thermal store. For
local heat transfer around each pile or borehole an innite line heat
source is assumed for short duration heat pulses. Globally and at
later times (dened as when the heat exchangers are thermally
interacting) a steady state is assumed within the store and subsequent heat input leads to linear changes in temperatures
throughout the store. The DST was initially validated against eld
data for small diameter (<50 mm) borehole thermal stores in
Sweden [25]. Subsequently, the DST approach has been implemented specically for use with pile heat exchangers in the software PILESIM [26]. PILESIM has been validated against eld data
[27], focussing on the overall heat exchange capacity of the system. Independent analysis using time-stepping nite element
models [28] implies that for regular arrays of piles the results
provided by PILESIM are appropriate. However, the DST assumes a
large number of identical piles installed in a regular array within a
circular plan area and it is not clear what errors result from smaller
or less regular pile group arrangements that are more representative of typical foundation layouts. As with line and cylinder source
based methods, the DST also assumes that the pile itself is at an
instantaneous steady state, employing constant resistance values to
account for the temperature difference between the uid and the
ground.
Numerical methods have also been used to assess multiple
ground heat exchangers, e.g. Refs. [29,30]; this has the advantage of
being able to assess bespoke geometries and arrangements of heat
exchangers. However three dimensional models of many heat exchangers can be computationally expensive and hence 2D simplications may be preferable. This, however, has the disadvantage of
not being applicable to long term conditions. One 2D model has
been used to test the applicability of using regular arrays of heat
exchangers as a simplied representation of irregular arrangements. This is particularly relevant for pile heat exchangers as the
pile positions are governed by the structural layout of the building

and not usually installed on a grid pattern. Teza et al. [31] concluded
irregular patterns can be replaced by regular ones for analysis with
a temperature prediction accuracy of 1  C, provided that the
average spacing and plan area are equivalent between the two
cases.

2.3. Heat exchanger conguration

The heat transfer pipes from multiple heat exchangers can be
connected in series, in parallel or in a combination of both. This detail
has an impact on the temperatures that develop around each individual heat exchanger. For pile heat exchangers connected in parallel
the inlet temperature to each pile (Ti-in) will be equal to the outlet
temperature from the heat pump (Thp-out). If all n piles in the system
are of the same geometry then each pile will also be subject to the
_ hp =n, and assuming that the heat transfer to each
same ow rate m
is also equal, the return temperature to the heat pump (Thp-in) will be
equal to average outlet temperature (Ti-out) from all of the piles:

Pn
Thpin

Tiout
n

(3)

If the design is based on a specied total thermal input Q nHq

(in W), then for n piles of length H, it is also true that:



_ hp Thpin  Thpout
Q Sc m

(4)

_ hp is the mass
where Sc is the specic heat capacity (J/kg K) and m
ow rate (kg/s) of the heat transfer uid at the heat pump. The Gfunction for the multiple heat exchangers, calculated by superposition, is used to determine the mean uid temperature for the
heat exchangers, Tf, where:

Pn
Tf

1 Tiout

Tiin

2n

(5)

The inlet and outlet temperatures to the heat pump can then be
determined as follows (where positive thermal loads represent
heat injection to the ground):

Thpout Tf Q=2S m
c _ hp

(6)

Thpin Tf  Q=2S m
c _ hp

(7)

F. Loveridge, W. Powrie / Energy 64 (2014) 747e757

This calculation clearly depends on the important assumption

that the heat ux q to each pile is the same. Eskilson suggests that
this simplication is valid for heat exchangers in parallel, with
discrepancies to more detailed calculations being of the order of
0.1  C [11]. He also suggests that the error would be larger if the
heat exchangers are coupled in series or if there is a variation of q
along their length, although the size of this error is not specied.
If the piles are connected in series then all piles will be subject to
the same uid ow rate as the heat pump. The inlet temperature to
each pile will always be equal to the outlet temperature of the
preceding pile in the circuit. Both the temperature difference between pile inlet and outlet and the heat ux to each pile will now be
different. The length of the series circuit of piles will determine how
big the errors are when using Equations (3)e(7) above. In reality
most systems consist of a combination of series and parallel circuits
to optimise the ow conditions. Calculation of the exact inlet and
outlet temperature to each individual pile is therefore very time
consuming. Fortunately Katsura et al. [22] have shown that simplications, including assuming each series circuit acts as one heat
exchanger, provide sufciently accurate results. Providing the piles
are all the same geometry, this is effectively the same as assuming
all individual heat exchangers operate in parallel. While more
complicated approaches can be adopted, and it has been shown
that maximum energy output can be obtained by adjusting the ow
rate and position of each pile individually [32,33], this complexity is
considered beyond the scope of the current paper. Instead we focus
on providing the rst multiple heat exchanger G-functions specifically for pile heat exchangers and investigating how typical piled
foundation arrangements can affect energy efciency.

Table 1
Main characteristics of 3D pile G-function model, see also Ref. [5].
Pile geometry
Model extent
Heat transfer
Initial conditions
Boundary conditions

Elements
Mesh sizes
Material properties
Validity
Software used

As Fig. 1, but with AR 15, 25, 33 1 3, 50

To a distance of 25 m radially and below the pile;
quarter pile modelled for computational efciency
By conduction only; pipes and uid not modelled
as these reach a steady state rapidly
Constant temperature (T 0  C) applied throughout
Constant heat ux applied at pipe outer radius
Constant temperature (T 0  C) applied at far
eld boundaries and ground surface
Insulated boundaries on the planes of symmetry
8 Node linear heat transfer brick elements
5 mm at pipes; 10 mm at pile edge; 2.5 m at
far eld; 0.5 m along pile length
Refer to Fig. 1
Fo  200
ABAQUS 6.10-2
=

750

3. Multiple pile G-functions: ground temperature response

3.1. Two pile G-functions

3.1.1. Basis
For Fo < 200 the 3D model used to develop the single pile Gfunctions was applied to determine revised functions for two piles
interacting. Full details of the model development and validation
are given in Ref. [5]; summary characteristics are given in Table 1.
For Fo > 200 nite line source methods are applicable. In this case
the nite line source was evaluated numerically on the basis on an
axisymmetric model constructed in ABAQUS. This allowed any pile
spacing to be easily assessed. The heating input was applied at the
position of the pile radius as shown in Fig. 2. Although this effectively makes the model a nite cylinder, this is equivalent to a line
source after a short time period and the model output was used
only for Fo > 200. Full details of the model are given in Ref. [14]
with summary characteristics provided in Table 2.

To calculate the G-function for two piles interacting the temperature change at r rb at the edge of a single pile was added to
the temperature change in the model at r B where B is the centre
to centre spacing between the two piles. As the pile arrangement is
symmetrical and both piles suffer the same degree of interaction
then the G-function only needs to be calculated for one pile. The
length of both piles is then taken into account when determining
the applied heat ux as q Q =2H. When presenting his G-functions, Eskilson normalised the heat exchanger spacing B by the
length of the heat exchanger, H [11]. However, with piles we have
found it more practical to normalise by the pile diameter 2rb, as a
result of existing practice in the piling industry. First, pile spacings

Table 2
Main characteristics of axis-symmetric nite line source model, see also Ref. [14].
Pile geometry
Model extent
Heat transfer
Initial conditions
Boundary conditions

Elements
Mesh sizes
Material properties
Validity
Software used

As Fig. 2, with rb 0.6 m and AR 15, 25, 33 1 3, 50

To a radial distance of 150 m and depth of 200 m
By conduction only; pile not modelled as applied
in long term only
Constant temperature (T 0  C) applied throughout
Constant heat ux applied at pile radius over the
full length of the pile, insulated below the pile toe.
Constant temperature (T 0  C) applied at far
eld boundaries and ground surface
4-Node linear axisymmetric heat transfer
quadrilateral elements
10 mm at pile edge; 5 m at far eld; 1 m along
pile length
As Fig. 1
Fo > 200
ABAQUS 6.10-2
=

Single pile G-functions have been presented by the authors in

Ref. [5]. These were derived by numerical analysis methods, using
two dimensional cross-sections through the pile for short timescales to determine lower and upper bound conditions and larger
three dimensional models to take into account the development of
a steady state in the longer term. The geometry and material
properties for the upper and lower bound piles are shown in Fig. 1.
Generally large diameter piles with pipes installed close to the pile
edge are represented by the upper bound, while large diameter
piles with pipes installed in the centre of the pile are represented by
the lower bound. Smaller diameter piles and those with pipes
installed in intermediate positions fall between the bounds. In
addition, piles where the concrete thermal conductivity (lc) is
greater than that of the ground (lg) tend towards the upper bound,
while the reverse conditions lg > lc tend towards the lower bound.

Fig. 2. Schematic of axis-symmetric nite line source model.

F. Loveridge, W. Powrie / Energy 64 (2014) 747e757

are often specied as a centre to centre spacing in terms of a

multiple of the diameter. Secondly, pile lengths are often determined by geological features, such as the presence of a hard stratum or the need to avoid penetrating into an underlying aquifer. In
addition, while Eskilson only published G-functions for AR 1000,
it is important to consider a range of smaller ARs for piles. Values of
15, 25, 33 1 3 and 50 have been used on the basis of a survey of
constructed geometries [6].
=

3.1.2. Results
Two-pile G-functions for the extreme cases of AR 15 and
AR 50 are plotted in Fig. 3 for different B/2rb values from 1.2 (the
closest piles are typically spaced) to 20 (where the inuence of
adjacent piles diminishes). The inuence of the pile spacing is clear,
with much greater temperature changes at steady state for those
piles at closest spacing. Table 3 gives an indication of the maximum
interaction experienced by the piles, measured as a percentage
increase in F in the long term. This clearly shows the greater
interaction for higher AR piles, but also the important inuence of
the spacing between the piles. For B/2rb 20, increase in F values
are always less than 15%. However, for B/2rb 1.2, F increases by up
to 76%. These increases in F mean that there are diminishing
returns available from multiple piles as their spacing decreases.
This means that for the closest spacing, less than 60% of the energy
that could be obtained from a single pile is available from each of
the pair of piles. Therefore in total 1.2q is available compared with
2q for two isolated piles. This means signicantly diminishing
returns as piles get closer together.
Fig. 3 and Table 3 also show that the higher aspect ratios have
greater values of F as a steady state develops, and also a greater
degree of interaction as seen by the bigger increase in F and corresponding reduction in equivalent energy for the two pile case.
Piles with a lower aspect ratio interact to a lesser extent due to the
earlier and more signicant inuence of the surface boundary
condition. This restricts both the overall temperature change that
can occur, and also the distance to which temperatures within the
ground are inuenced by the heat exchanger. This suggests that low

751

Table 3
Steady state increase in F output for a pair of piles for different spacings and aspect
ratios (AR).
B/2rb

20
10
5
3
2
1.5
1.2

AR 15

AR 33

AR 25

AR 50

Increase in F

Increase in F

Increase in F

Increase in F

3%
9%
23%
37%
49%
59%
67%

6%
15%
30%
44%
56%
65%
71%

8%
19%
34%
48%
59%
67%
74%

12%
25%
40%
52%
63%
70%
76%

aspect ratio piles have the potential to be more efcient due to i)

reduced temperature changes in the ground and ii) reduced interactions between adjacent heat exchangers.
Fig. 4 plots F vs Fo for different aspect ratios, showing that the
upper and lower bound cases experience identical interaction in
the long term for a given aspect ratio. However, in the short term
there is a small increase in interaction for the upper bound case.
This is to be expected given the greater proximity of the pipes to the
pile edge. Overall Fig. 4 shows that the interactions are similar at
small values of time regardless of the pile spacing and pile geometry. This is also to be expected as the zone of steep temperature
gradients has yet to reach far beyond the pile. However, as time
increases larger aspect ratio piles interact more and for longer.
Eskilson reported that there is no interference between borehole heat exchangers if B > H and that any interactions are small as
long as B > H/2 [11]. For the case of pile heat exchangers, the latter
criterion is equivalent to B/2rb > AR. Reference to Fig. 4 shows that
these guidelines remain true for pile heat exchangers and that in
this case small interactions are approximately equivalent to a 5%
increase in F. However, for pile heat exchanger systems, where the
pile layout and spacing are governed by the structural and
geotechnical design of the building and foundations, it would be
exceedingly unlikely for the pile spacing to be 15 m or more.
Therefore in most cases some interactions will occur.

Fig. 3. Pile G-functions for two piles interacting at different B/2rb values: a) AR 15 lower bound; b) AR 15 upper bound; c) AR 50 lower bound; d) AR 50 upper bound. In
each case the curves are, from top to bottom, B/2rb 1.2, 1.5, 2, 3, 5, 10, 20 and N respectively.

752

F. Loveridge, W. Powrie / Energy 64 (2014) 747e757

Fig. 4. Percentage increase in F for different normalised pile spacings and aspect ratios (AR): a) AR 15; b) AR 25; c) AR 33; d) AR 50. Solid lines for upper bound cases;
dashed lines for lower bound cases.

3.2. Multiple pile G-functions

Using Figs. 3 and 4 and Table 3 it is possible to interpolate the
inuence of piles at any spacing and hence compile the average
multiple pile G-function for any arrangement of piles. For brevity,
two examples are included below; three piles in a line (Fig. 5),
which may be an arrangement beneath a building column and 9
piles in a grid (Fig. 6), which gives an idea of the degree of inuence
of many pile heat exchangers. In reality most pile arrangements are
irregular and related to the structural layout of the building, in
particular under the positions of the columns that transfer most of
the vertical load to the foundations. A column may be supported by
two or three closely spaced piles, with many columns being present
at larger spacings. However, the following method may be used for
any arrangement.
It has been assumed, for the purpose of the examples, that the
pipes are installed near the edge of the pile and therefore an upper
bound pile G-function has been used. An aspect ratio of 50 has been
chosen because this provides an upper estimate to the likely interactions for pile heat exchangers. Curve tting was then applied
to Fig. 4d so that the percentage increase in F for any pile spacing
would be determined. For three piles in a line (Fig. 5) the G-function
will then be equal to the average response of the three piles, one of
which will suffer interactions from two of its neighbours at spacing
B and two of which experience interactions from two piles at B and
2B spacing respectively. The same procedure can be applied for the
nine pile arrangement (Fig. 6), except here there are 4 corner piles,
4 centre edge piles and one central pile, all of which have different
combinations of interactions which must be averaged.
Figs. 5 and 6 show the signicant impact that multiple pile
heat exchangers can have when interacting. For a single
pile (B innity) with AR 50, the long term steady state normalised temperature response (F) is approximately 3.6. For two
piles at B/2rb 1.2 the response increases to 6.3, for three piles and

Fig. 5. G-Function for three upper bound piles in a line, with AR 50. From top to
bottom B/2rb 1.2; 1.5; 2; 3; 5; 10; 20; N.

nine piles at the same spacing it is 8.6 and 22.2 respectively. These
are large increases and will result in a corresponding decrease in
the energy that can be exchanged per linear metre of the pile.
Table 4 summarises the increase in F for the closest spacing. These

Fig. 6. G-Function for nine upper bound piles arranged on a grid, with AR 50. From
top to bottom B/2rb 1.2; 1.5; 2; 3; 5; 10; 20; N.

F. Loveridge, W. Powrie / Energy 64 (2014) 747e757

Table 4
Steady state increase in F energy output for different pile arrangements compared to
a single pile (AR 50).
No piles

Spacing

Increase in F

Single pile
2 Piles
3 Piles in a line
9 Piles in a grid

N/A
B/2rb 1.2
B/2rb 1.2
B/2rb 1.2

N/A
76%
141%
521%

increases mean that at steady state with nine piles in a grid that
each pile is only delivering 16% of the energy of an individual isolated pile. However, it is rare for so many piles to be spaced so
closely and in reality combination arrangements as described above
are more common. In addition, this analysis has assumed that all of
the nine piles are equipped as heat exchangers. The results show
that while each additional pile used as a heat exchanger does increase the overall quantity of energy available there are diminishing returns. When the energy required for uid circulation is also
taken into account it may be more economical to equip only some
of these piles with heat transfer pipes.

4. Multiple concrete G-functions: internal pile temperature

response
The multiple pile G-functions presented in Section 3 describe
the average temperature change with time in the ground around
the heat exchangers. However, it is important to determine
whether the nature of the temperature changes within the pile is
also affected by the interaction of the heat exchangers. To our
knowledge this has not been ascertained before. Previously the
authors have presented concrete G-functions to describe the transient behaviour of the pile concrete [5]. This is an improvement on
the previous approach of assuming the pile concrete is at steady
state. The rst concern for multiple piles is whether the change in
temperature eld for two or more piles will inuence the steady
state value of the pile thermal resistance. Secondly, we need to
understand whether the shape of the concrete G-function,
expressed as a proportion of the steady state resistance Rc with
time, changes when there are more than one pile in proximity.
To answer these questions, the numerical model used in Ref. [5]
was extended to represent two piles, taking advantage of a line of
symmetry between the piles to minimise the computational effort
required (Fig. 7). Full details of the model set up and validation are
given in Refs. [5,14], with updated geometry and a summary of the
conditions analysed provided in Table 5. Only the upper and lower
bound pile concrete conditions, as identied in Ref. [5] and given in

753

Table 5
Main characteristics of 2D concrete G-function model, see also Refs. [5,14].
Pile geometry
Model extent
Heat transfer
Initial conditions
Boundary conditions

Elements
Mesh sizes
Material properties
Validity
Software used

Upper and lower bound pile geometries as per Table 6

To a distance of 25 m from the model centre (Fig. 7)
By conduction only; pipes and uid not modelled
as these reach a steady state rapidly
Constant temperature (T 0  C) applied throughout
Constant heat ux applied at pipe outer radius
Constant temperature (T 0  C) applied at model edge
Insulated boundary at plane of symmetry
Triangular
2 mm at pipes; 10 mm at pile edge; 0.5 m model edge
Refer to Table 6
Fo  10
COMSOL 4.3

Table 6, were analysed. These include upper and lower bounds for
the two cases of pipes installed centrally and those closer to the pile
edge. As the lower bound piles are 300 mm in diameter with two
pipes installed, it is necessary to compare two cases. Case #1 is
where the two pipes are aligned perpendicular to the line of
symmetry in the model (as shown in Fig. 7) and Case #2 is where
the pipes are aligned parallel to the line of symmetry. It should also
be noted that lower and upper bound cases used for the concrete Gfunction are different from the lower and upper bound cases used
for the pile G-functions as they relate to the behaviour of the
concrete not of the ground.
4.1. Steady state resistance for two interacting piles
At any given time the thermal resistance of the concrete part of
the pile is calculated as follows:

Rc

Tp  Tb
q

(8)

where Tp and Tb are the integral mean values of the temperature at

the pipes and the pile edge boundaries respectively, and q is the
total heat ux (in W/m) applied to the all the pipe boundaries. The
steady state resistance is the asymptotic value Rc calculated at
larger values of time, typically when Fo approaches 10. Table 6
presents the steady state values of Rc for single and pairs of interacting piles. In most cases the presence of an additional pile has no
noticeable impact on the steady state resistance, despite some
changes to the heat ow paths that the additional pile must cause.
The exception to this is the lower bound case with pipes near to the
edge. Here the steady state resistance now depends on the relative
arrangement of the two pipes in the adjacent piles. For Case #1 the
resistance is increased by 2%, while for Case #2 it is decreased by
2%. These differences are not signicant in the context of construction tolerances for the positioning of the pipes within the
piles. In addition, the construction process will give limited control
over the orientation of any pair of pipes, and therefore the nal
orientation will be almost random. Therefore, overall, it is considered appropriate to continue using the steady state resistance of a
single pile regardless of the number and spacings of adjacent piles.
4.2. Concrete G-functions for two interacting piles

Fig. 7. Schematic of numerical model for 2D analysis of concrete resistance and

temperature response (showing Case #1, not to scale). Refer also to Table 5.

The second question that needs to be answered is whether the

shape of the concrete G-function curve, i.e. how the transient value
of thermal resistance for the pile changes with time, is affected as a
result of piles interacting. Fig. 8a and b plots concrete G-functions
for upper and lower bound cases for the scenario of pipes placed
centrally and near the edge respectively. In both cases the solid
lines represent the G-function for a single pile and the dashed lines

754

F. Loveridge, W. Powrie / Energy 64 (2014) 747e757

Table 6
Steady state thermal resistance (Rc) for individual and pairs of interacting piles.

Central pipes;

Pipes near edge;

Single pile
Two piles#1
Two piles#2
Single pile
Two piles#1
Two piles#2

Lower bound
300 mm diameter
pile, 2 pipes
lc 1; lg 2

Upper bound
1200 mm diameter
pile, 4 central pipes
or 8 near edge
lc 2; lg 1

0.267
0.267
0.267
0.183
0.187
0.180

0.231
0.231
N/A
0.029
0.029
N/A

m
m
m
m
m
m

K/W
K/W
K/W
K/W
K/W
K/W

m K/W
m K/W
m K/W
m K/W

For piles with centrally placed pipes the concrete cover is c 105 mm for 2rb 300
mm and c 555 m for 2rb 1200 mm. For piles with pipes placed near the edge the
concrete cover is c 50 mm for 2rb 300 mm and c 75 m for 2rb 1200 mm

represent those for two piles interacting. There is some difference

between these cases, but typically the discrepancies are within a
few percent. The exception is the lower bound case with pipes near
the edge where the difference between the single and two pile
scenarios is initially 20%, rapidly falling to around 2% by Fo 0.1.
Given that the lower bound represents smaller diameter piles and
that Fo 0.1 is equivalent to less than 1 h for a 300 mm diameter
pile, this difference is minor in the context of standard 1 h time
stepping used in design. Therefore, while it is possible to produce
new concrete G-functions for the interacting cases in Fig. 8, it will
be sufcient in most scenarios to use the curves for single piles
presented (with curve t data) in Ref. [5].

Fig. 9. Example thermal loads for one year commencing in January (insets show daily
cycle detail). Note: positive thermal loads are heat injection to the ground (building
cooling); negative thermal loads are heat extraction from the ground (building heating).

describe the temperature change in the ground, while a lower

bound concrete G-function for pipes placed near the edge is used to
determine the temperature change within the pile. The following
equation is used to calculate the mean temperature change in the
heat transfer uid:

DTf qRp qRc Gc

q
Gg
2plg

(9)

where Gc is the concrete G-function and Gg is the pile G-function. As

the applied thermal load q (in W/m) is not constant, but changes
with each time step, superposition must be used with each Gfunction so that the change in temperature can be calculated:

5. Impact on energy storage

To illustrate the impact of interactions between multiple pile
heat exchangers on the thermal energy that can be extracted from
the ground, three cases have been compared using the multiple pile
G-functions presented in Section 3:
 A single pile (B/2rb innity in Fig. 3d),
 Two piles installed at a spacing of B/2rb 1.2 (Fig. 3d),
 Nine piles installed on a grid pattern with B/2rb 3 (Fig. 6).
In all cases AR 50 has been assumed as this represents the
worst case interactions. The piles have been selected to be 450 mm
diameter and 22.5 m long with a concrete thermal conductivity of
lc 1 W/mK. The surrounding ground is assumed to have thermal
properties lg 2 W/mK and ag 1  106 m2/s. The pile concrete
thermal resistance Rc is 0.075 mK/W and the pipe resistance Rp is
0.025 mK/W for four pipes placed near the edge of the pile.
Consequently an upper bound pile G function for AR 50 is used to

DTn


i
qi h 
G Fon  Foi1  GFon  Foi
pl
2
g
i1

iX
n

(10)

where n is the point in normalised time in which the superposition

is evaluated and G is the G-function (whichever one is being used in
a particular case) calculated at the value of Fo prescribed in the
equation. Equation (10) has been coded in the software Matlab to
allow calculation of the sum in hourly timesteps for the period of
one year.
For the examples being considered q is obtained from the
building thermal load prole shown in Fig. 9. The prole is given for
a typical year and has been developed from a numerical simulation
of a modern multi-use development in the South East of England,
scaled down to an appropriate level for a single pile. The total demand is 1.72 MWh heating and 1.76 MWh cooling per pile. However, despite the overall energy demand being close to balanced,
the peak power values for cooling are much higher than those for

Fig. 8. Concrete G-functions for individual and pairs of interacting piles: a) pipes placed centrally; b) pipes placed near the pile edge. Solid lines are for individual piles; short dashed
lines are two piles Case #1; long dashed lines are two piles Case #2.

F. Loveridge, W. Powrie / Energy 64 (2014) 747e757

755

Fig. 10. Calculated mean uid temperatures for different numbers and arrangements of piles.

heating, being comprised of shorter duration but greater magnitude power peaks. It will be seen that this has an impact on the
resulting temperature changes.

5.1. Results
The calculated mean uid temperatures (Tf) are shown in Fig. 10.
The actual inlet and outlet temperatures to the heat pump would
actually cover a wider range than this (as per Equations (6) and (7)).
However, these values have not been calculated as they would
depend on the mechanical design of the pipework for any particular system, which would need to be optimised to maintain a
sensible temperature differences across the heat pump. Fig. 10
shows the calculated temperature changes for the rst year,
which range from 8.3  C to 15.9  C for single pile to 11.1  C
to 20.1  C for two piles at B/2rb 1.2 and 11.0  C to 21.8  C for
nine piles at B/2rb 3. It is interesting to note that initially the two
pile scenario has a greater temperature change that the nine pile
scenario. This is due to the closer spacing and the interaction effects
commencing at a shorter time. In the longer term the temperature
change for the nine pile arrangement is much greater, especially in
the summer when the rate of change of thermal load is greatest. It

is expected that the differences between the three cases would

increase with time.
Assuming a typical initial ground temperature of 12  C it is
reasonable to apply limits on the mean uid temperature change
of 10  C and 20  C (i.e. absolute limits of 2  C and 32  C). Then
the average energy available per pile can be calculated using
Equation (9) assuming that DTf remains within the range of these
limits. As in Ref. [5], for simplicity, the shape of the thermal load
proles in Fig. 9 has not been changed, but the values of applied
thermal load have been adjusted pro-rata to give an indication of
the impact the interactions are having on the thermal capacity of
the piles. In reality the system would be adjusted to cover as much
base load as possible with some extreme peak loads being be
supplied by an auxiliary system. The assessment would also cover
the full design life of the structure rather than just the rst year as is
shown here.
Fig. 11 shows the relationship between the average energy
available per pile and the imposed temperature limits for the one
year period analysed. Based on the 10  C temperature change
limit the available average heating energy per pile is 2069 kWh,
1546 kWh and 1564 kWh for one pile, two pile and nine piles
respectively. Interestingly the nine pile case is actually slightly
greater than the 2 pile case. This is because the nine piles are at

Fig. 11. Relationship between allowable uid temperature change and available energy for example pile arrangements: a) heating; b) cooling.

756

F. Loveridge, W. Powrie / Energy 64 (2014) 747e757

greater spacing and the analysis starts in the winter with heating. If
the analyses were carried out over a longer time period, there
would undoubtedly be less energy available for the nine pile case
than for the 2 pile case. In terms of cooling, using the 20  C limit,
the available average energy for each pile is 2219 kWh, 1746 kWh
and 1611 kWh for one pile, two pile and nine piles respectively. The
range of values is similar to the heating case, but as the cooling
season comes later in the analysis period there is a greater difference between the two pile and nine pile cases. Despite the similar
amounts of energy extracted in heating and cooling the temperature changes are bigger in cooling due to the higher peak loads. This
illustrates the importance of understanding both the monthly and
total energy demands of a systems and the shorter term variations
in demand which will result in the peak loads.
In total the average energy available per pile in the three arrangements is 4288 kWh, 3292 kWh and 3175 kWh for the single
pile, two pile and nine piles cases respectively. This represents a
23% drop in available energy from one to two piles and a further
26% drop for the nine pile case. It is interesting that these two
gures are of the same order, despite a much larger number of piles
in the nine pile case. This reects the importance of pile spacing
and the much reduced interactions when the pile spacing is opened
from B/2rb 1.2 to B/2rb 3. The equivalent energy for the nine pile
arrangement is 51% compared with 23% for that arrangement at
steady state. This indicates that there is the potential for the efciency of the system to decline further over the lifetime of the
building. For the 450 mm diameter piles analysed one year represents Fo 623. Steady state conditions (under constant thermal
load) are not reached until in excess of Fo 10,000, or around 16
years.
6. Discussion
Ground heat exchangers installed in multiple piles will interact
adversely in terms of energy available per pile, generally by less
than 5% as long as B/2rb > AR. However, given the cost of constructing deep foundations, priority will always be given to optimising pile layouts with respect to their structural function, that of
supporting the overlying building. This means that it is not possible
to adjust the positions of pile heat exchangers to increase the
spacing and hence maximise the energy output. However, an
assessment of the potential for interactions (increased temperature
change or reduce energy output) between adjacent piles should
still be carried out.
The example analyses presented have shown how the effect of
adjacent piles can greatly reduce the energy output from each individual pile. Overall the total energy obtained from multiple piles
is always greater than from a single pile, but as the number of piles
is increased and the spacing reduced the energy return per pile
decreases. In some extreme cases it may be more economical to
equip only some piles in a foundation layout with heat exchange
pipes. While the pipes themselves are of relatively low cost, there is
additional programme time for installing the pipes and additional
running costs for a longer pipe circuit. These would need to be
weighed against the energy gains from the interacting piles.
The degree of interactions between adjacent pile heat exchangers will depend on a number of factors. The spacing of the
piles is important, but so is the number of piles. The spacing will
have the biggest impact on the time taken for interactions to
become signicant, while the number of piles in the arrangement
(in combination with their spacing) will impact the long term energy obtainable. Generally smaller aspect ratio piles will interact to
a lesser extent than larger aspect ratio piles, which means that the
former can be successfully implemented at closer spacings. The
nature of the thermal load is always important for the performance

of ground heat exchangers and this case is no exception. Short term

variations in load will reduce interactions compared with a sustained base load.

7. Conclusions
This paper presents a method for determining new G-functions
for use in the thermal analysis of multiple interacting pile heat
exchangers. Example pile G-functions, which describe the temperature change in the soil around a pile with time, are presented
for a number of examples congurations. The key conclusions of
this study are:
 If multiple adjacent piles are used as heat exchangers then there
will be adverse thermal interactions between the heat exchangers. These interactions will become more signicant over
the lifetime of the energy system, but will also be dependent on
the nature of the thermal load. For example highly uctuating
thermal demands will reduce the potential for interactions between heat exchangers.
 Heat exchangers with smaller aspect ratios are affected less by
thermal interactions. Thus interactions between piles will be
less than those between traditional borehole heat exchangers
installed at the same spacing, potentially leading to more energy
efcient systems.
 However, as pile spacing is usually governed by the overlying
structure, piles are generally more closely spaced than typical
borehole arrangements.
 Consequently, it is not always advantageous to equip all piles
within a pile group with heat transfer pipes.
 The degree of interactions to be expected for a given scheme can
be calculated using the methods for multiple pile G-functions
described in this paper.
 Signicantly, it was found that the changes to the temperature
eld within concrete piles are not sufcient to cause appreciable changes to either the transient or steady state resistance
of the pile.

References
[1] Lund JW, Freeston DH, Boyd TL. Direct utilization of geothermal energy: 2010
worldwide review. Geothermics 2011;40(3):159e80.
[2] Hamada Y, Saitoh H, Nakamura M, Kubota H, Ochifuji K. Field performance of
an energy pile system for space heating. Energy Build 2007;39:517e24.
[3] Gao J, Zhang X, Liu J, Li K, Yang J. Thermal performance and ground temperature of vertical pile foundation heat exchangers: a case study. Appl Therm
Eng 2008;28:2295e304.
[4] Bradl H. Energy foundations and other thermo active ground structures.
Geotechnique 2006;56(2):81e122.
[5] Loveridge F, Powrie W. Temperature response functions (G-functions) for
single pile heat exchangers. Energy 2013;57:554e64.
[6] Loveridge F, Powrie W. Pile heat exchangers: thermal behaviour and interactions. Proc ICE Geotech Eng 2013;166(2):178e96.
[7] Li M, Lai ACK. New temperature response functions (G-functions) for pile and
borehole ground heat exchangers based on composite-medium-line-source
theory. Energy 2012;38:255e63.
[8] Man Y, Yang H, Diao N, Liu J, Fang Z. A new model and analytical solutions for
borehole and pile ground heat exchangers. Int J Heat Mass Transf 2010;53:
2593e601.
[9] Hellstrom G. Ground heat storage, thermal analysis of Duct storage systems,
theory. Sweden: Department of Mathematical Physics, University of Lund;
1991.
[10] Lamarche L, Beauchamp B. A new contribution to the nite line-source model
for geothermal boreholes. Energy Build 2007;39(2):188e98.
[11] Eskilson P. Thermal analysis of heat extraction boreholes. Doctoral Thesis.
Lund, Sweden: University of Lund, Department of Mathematical Physics;
1987.
[12] Ingersoll LR, Zobel OJ, Ingersoll AC. Heat conduction with engineering and
geological applications. 3rd ed. New York: McGraw-Hill; 1954.
[13] Bernier M. Ground coupled heat pump system simulation. ASHRAE Trans
2001;107(1):605e16.

F. Loveridge, W. Powrie / Energy 64 (2014) 747e757

[14] Loveridge F. The thermal performance of foundation piles used as heat exchangers in ground energy systems. Doctoral Thesis. Southampton, UK: University of Southampton, Faculty of Engineering and The Environment; 2012.
[15] Loveridge F, Powrie W. Performance of piled foundations used as heat exchangers. In: 18th international conference for soil mechanics and geotechnical engineering, Paris, France, September 2-5, 2013 2013.
[16] Hellstrm G, Sanner B, Klugescheid M, Gonka T, Mrtensson S. Experiences
with the borehole heat exchanger software EED. In: Proceedings of MEGASTOCK 97, Sapporo, Japan 1997. pp. 247e52.
[17] Spitler JD. GLHEPRO e a design tool for commercial building ground loop heat
exchangers. In: Proceedings of the fourth international heat pumps in cold
climates conference, Aylmer, Qubec. August 17-18, 2000 2000.
[18] Wood CJ, Liu H, Riffat SB. An investigation of the heat pump performance and
ground temperature of a pile foundation heat exchanger system for a residential building. Energy 2010;35(12):4932e40.
[19] Wood CJ, Liu H, Riffat SB. Comparison of a modeled and eld tested piled
ground heat exchanger system for a residential building and the simulated
effect of assisted ground heat recharge. Int J Low Carbon Technol 2010;5(12):
137e43.
[20] Arup. DTI partners in innovation 2002, ground storage of building heat energy.
Overview Report, O-02-ARUP3, May 2005. Available from:, http://www.arup.
com/_assets/_download/download352.pdf; 2005 [downloaded 13/03/2013].
[21] Katsura T, Nagano K, Takeda S. Method of calculation of the ground temperature for multiple ground heat exchangers. Appl Therm Eng 2008;28:1995e
2004.
[22] Katsura T, Nagano K, Narita S, Taekda S, Nakamura Y. Calculation algorithm of
the temperatures for pipe arrangement of multiple ground heat exchangers.
Appl Therm Eng 2009;29:906e19.
[23] Hellstrom G. Duct ground heat storage model, manual for computer code.
Sweden: Department of Mathematical Physics, University of Lund; 1989.

757

[24] Claesson J, Hellstrom G. Model studies of duct storage systems. In:

Millhone JP, Willis EH, editors. Proceedings of the international conference on
new energy conservation technologies and their commercialisation, 6th e
10th August, 1981. Berlin: International Energy Agency; 1981.
[25] Hellstrom G. Comparison between theoretical models and eld experiments
for ground heat systems. In: Proceedings international conference on subsurface heat storage in theory and practice, June 6 e 8, Stockholm, vol. 2.
Swedish Council for Building Research; 1983.
[26] Pahud D. PILESIM2, simulation tool for heating/cooling systems with heat
exchanger piles or borehole heat exchangers. User Manual. Lugano,
Switzerland: Scuola Universitaria Professionale della Svizzera Italiana; 2007.
[27] Pahud D, Hubbuch M. Mesures et optimisation de linstallation avec pieux
energetiques du Dock Midfoeld de laeroport de Zurich. Swiss Federal Ofce of
Energy; 2007.
[28] Markiewicz R. 2010. Pers. Comm.
[29] Fan R, Jiang Y, Yao Y, Ma Z. Theoretical study on the performance of an integrated ground source heat pump system in a whole year. Energy 2008;33:
1671e9.
[30] Zanchini E, Lazzari S, Priarone A. Long-term performance of large borehole
heat exchanger elds with unbalanced seasonal loads and groundwater ow.
Energy 2012;38:66e77.
[31] Teza G, Galgaro A, De Carli M. Long term performance of an irregular shaped
borehole heat exchanger system: analysis of real pattern and regular grid
approximation. Geothermics 2012;43:45e56.
[32] de Paly M, Hecht-Mendez J, Beck M, Blum P, Zell A, Bayer P. Optimization of
energy extraction for closed shallow geothermal systems using linear programming. Geothermics 2012;43:57e65.
[33] Beck M, Bayer P, de Paly M, Hecht-Mendez J, Zell A. Geometric arrangement
and operation mode adjustment in low-enthalpy geothermal borehole elds
for heating. Energy 2013;49:434e43.