Professional Documents
Culture Documents
Article v3
Article v3
Abstract
Resumindo... blalblablal
Keywords: Earth-air heat exchangers, keyword two
∗
Corresponding author
1. Methodology
1.1. Settings
In general, this work aims to evaluate how the EAHE performance is
affected by coupling the ducts to galvanized structures. In particular, our
simulations adapt ideas from the reference [1], regarding the climate and soil
conditions from an experimental EAHE installation in the south Brazilian
city of Viamão [2].
Here, we consider the straight part of the buried duct, as done in other
references like [3, 4, 5]. Moreover, in the duct longitudinal direction, the
soil temperature variations are considered negligible, relative to the ones in
transverse directions. Therefore, we estimate soil temperatures with a 2D
modeling approach, adopting computational domains similar to the sketch
2
in Fig. 1 (a). The overall strategy is to surround the duct by a galvanized
block which can be connected to fins made of the same highly conductive
material.
(a)
(b)
In the example of Fig. 1 (a), z0 is the duct center depth, initially equal to
1.6 m; D0 = 11 cm is the diameter of the duct; b0 = 18 cm refers to the block
size (which has a square cross section). Moreover, all fins have the same
width Sv = 1 m and thickness Sh = 1 cm. Here, we placed the duct and
block equally centered in the horizontal direction. Hence, their centers have
the coordinate x0 = 5m. Figure 1 (b) also gives a 3D view, where we can
see that the duct, block and fins have the same length L0 = 25.77 m. Such
values for z0 , D0 , and L0 are used to emulate the experimental installation
3
from [2]. As explained in the Section 1.4, the values for the galvanized parts
represent an adaptation of the work [1] aimed to the reality given in [2].
The thermophysical properties of the soil and air in Viamão, as well as
the galvanized materials, are given in Table 1. The values come from the
references [6] and [1]. As done in other works like [7, 4, 8], this study does
not consider the thickness and properties of the ducts; besides, references
like [9] conclude that they have little influence on thermal exchanges because
the thickness of the ducts is relatively small.
UL = πD0 h (2)
1
German for graphical design of EAHE
4
is the overall heat transfer coefficient (per duct length) between the air stream
and the duct walls.
We approximate the convection coefficient at the inner surface of the duct
with
λa Nu
h= , (3)
D0
where the Nusselt number is determined by
Nu = 0.0214 (Re0.8 − 100) Pr0.4 . (4)
The Reynolds and Prandtl numbers are given, respectively, by
ρa va D0
Re = , (5)
µa
µa cp,a
Pr = . (6)
λa
The soil temperature Ts is estimated disregarding the ducts presence. At
this point, we assume that it is already known, but we explain ahead how we
computed it adicionar seção. Hence, GAEA needs to estimate a corrected
temperature on the duct walls, i.e.
U ∗ Ts + Ta
Tc,w = . (7)
U∗ + 1
Here, Tc,w is a weighted average between the air temperature inside the duct,
Ta , and the soil temperature on the duct wall (not influenced by its presence),
Ts . Therefore, U ∗ works as a weighting factor.
However, instead of calculating Tc,w directly by Eq. (7), the duct is divided
into 100 segments, where the thermal exchanges are determined iteratively
[10]. The idea is to take sufficiently small pieces to assume that the air flows
through them at an approximately constant temperature. Therefore, jumps
in temperature values occur between distinct segments of size ∆y = L0 /100.
For each segment k, a correction of the soil temperature in the duct wall
is calculated using the formula
k
U ∗ Ts + Ta,i
k
Tc,w = . (8)
U∗ + 1
k
Here Ta,i is the air temperature at the inlet of the k-th segment, where the
heat transfer rate is given by
k k
Q̇k = UL (Tc,w − Ta,i )∆y. (9)
5
Such rate is also given by
k k
Q̇k = ṁ cp,a (Ta,o − Ta,i ), (10)
k
where Ta,o is the air temperature at the outlet of the k-th segment and ṁ
the air mass flow. As we are considering a circular duct, then
πD02
ṁ = ρa va
. (11)
4
k
Combining equations (9) and (10), we can isolate Ta,o to obtain
k k
k k
∆y UL Tc,w − Ta,i
Ta,o = Ta,i + . (12)
ṁ cp,a
In summary, we calculate with Eq.(12) the air temperature at the outlet
of each segment based on the air temperature at the inlet of it. We note
that the outlet of segment k is the inlet of the k + 1 one. When the iterative
process starts, we know the temperature at the inlet of the first segment
(k = 1) because it is also the duct inlet. The algorithm ends by finding the
temperature at the the last segment outlet (k = 100), which coincides with
the duct outlet.
Juntar em algum lugar informações sobre todas as implementações... To
make the simulations with the GAEA model, we implemented an in-house
code in the Matlab programming language.
1.3. Physical Model
In this work, we compute the temperatures in the soil, and in regions
covered by galvanized materials, disregarding the presence of the ducts. In
particular, we consider cases similar to the one illustrated by Fig. 1 (a),
therefore, we will neglect the temperature variations in the y direction, and
solve the energy conservation equation
2
∂ 2T
∂T λs ∂ T
= + 2 , (13)
∂t ρs cp,s ∂x2 ∂z
restricted to two-dimensional domains in the plane x − z.
Based on the references (REFS), we can assume the adiabatic boundary
conditions
∂T
= 0◦ C/m at z = 15m. (14)
∂z
∂T
= 0◦ C/m at x = 0 m and x = 10 m. (15)
∂x
6
To keep the continuity of the thermal fluxes on the interfaces Γ between
soil (s) and galvanized parts (g), we impose that
T = Ta at z = 0 m. (17)
For this work, we fitted the annual air temperature data in Viamão by least
squares, as done in other references like [12]. In our model, Ta is given by
the periodic function
2π
Ta (t) = 20.49 + 5.66 sin t − 5.30 , (18)
365
The tests showed that this initial condition affects only the first two
months of simulation. Hence, we ran simulations covering a total time of
one year and two months, then discarded the first two. This approach is
advantageous because it reduces the computational cost to evaluate a year
of EAHE operation. To make a comparison, in references like [14], where the
authors assume that the initial condition is equal to the annual mean soil
temperature, it affects the solution for almost one year. That means that
they needed to simulate two years and discard entirely the first one.
7
1.4. Numerical approach
The initial-boundary value problem, composed by Eq. (13) to (19), was
solved numerically. For the time discretization, we used finite differences,
more specifically, the first-order implicit Euler method [15, 16]. The finite
element method of Galerkin [17, 18, 19] was employed in the spatial dis-
cretization. We made all the simulations utilizing in-house codes, developed
in the Matlab software language. The codes cover the numerical methods
and the GAEA model.
To explore computational domains like the ones in Fig. 1, the meshes were
composed by the following 6 subdomains: soil (1), block (2), left fin (3), right
fin (4), upper fin (5) and lower fin (6); this structure is illustrated in Fig. 2.
Hence, we could use the same mesh for different cases. For example, in the
case of a simulation involving the four fins, we could define that subdomains
2 to 6 had the thermal diffusivity of galvanized material, while subdomain
1 had thermal diffusivity of the ground. On the other hand, in another
simulation where the left fin is disregarded, then, we could just define the
thermal diffusivity of subdomain 3 to be equal to the soil diffusivity. The
results for such cases are shown ahead in colocar a seção.
8
In all mesh tests, we compared the annual temperature solution vectors
at the center of the block, i.e., at the coordinates x = 5 m and z = 1.6 m.
Using the infinite norm2 , we compared the differences between two successive
solution vectors T s . The conclusive tests and comparisons are presented in
Table 2, which gives also the number of nodes, elements and total simulation
time for each of them. As it can be seen in the same table, the results
converged as expected with the increase in the number of nodes and mesh
elements. Following [12], a maximum difference of the order of 10−4 between
two solutions was considered sufficient to complete the comparisons.
Table 2: Mesh tests
2
The infinite norm of a vector x = (x1 , · · · , xn ) is defined by: ||x||∞ = maxi=1,...,n |xi |
[21]
9
Figure 3: Selected mesh with 3935 elements
10
we follow the references [22, 14, 12] and measure the EAHE annual efficiency
by qR
365 2
0
Pt dt
θ = qR . (23)
365 2
0
Ps dt
In this work, all the results for the potentials are fitted by least squares
to periodic functions, like
11
based on conservation equations that are solved by finite volume methods
with the Fluent software.
Hence, Fig. 4 presents a comparison among: (V) the discrete, daily av-
erage temperatures at the outlet of one EAHE installation from [7]; (VF)
a least squares fitted curve of the same data, following a periodic function
similar to that given in Eq. (24); (D) and (H) the numerical results from the
models of [12] and [4], respectively; (CM) our current model results.
28
24
T (oC)
20
V.
16 V.F.
D.
C.M.
H.
12
0 50 100 150 200 250 300 350
t (days)
Figure 4: Comparison among the current model results againt others from literature
As we can see from Fig. 4, the current model gives results approximatelly
equal to the ones from [12]. The Pearson correlation coefficient between such
models results and the experimental data is about 0.89. Furthermore, the
annual RMS difference between the model results and the fitted experimen-
tal values is approximately 1.45◦ C. In [12], the authors estimated that the
correlation between the model results from [4] and the experimental data are
approximately 0.92, while the annual RMS difference between their results
and the fitted experimental values is close to 1.58◦ C.
Therefore, we concluded that our methodology is valid, providing accurate
and highly correlated results to the experimental ones. Furthermore, it can
12
be seen that the methodology is computationally efficient, as the simulations
can be performed in a few minutes (see Table 2), instead of hours, as it
happens with the complete 3D model from [4].
cabe falar um pouco das limitações do GAEA, sugestão Elizaldo.
(a) Summer
(b) Winter
Figure 5: Daily average soil temperature fields in the summer and winter
13
thermal potential is little more than 5◦ C.
2. Results
2.1. Introducing the block and fins
Using the mesh of Fig. 3, we tested twelve different configurations for the
block and fins coupled to a duct placed at the depth of z0 = 1.6 m. All
cases are illustrated in Fig. 6, and they are numbered from 1 to 12. The case
shown in the validation (where there is no block and no fins) is considered
a reference case, or case number 0. The article [1] considered 7 possible
arrangements, which we adapted here with the cases 0, 2, 3, 4, 5, 6, and 12.
By simetry, the case 4, corresponding to use a fin on the right side of the
block, is equivalent to use a fin on the left.
Table 3 presents the annual RMS values for the soil thermal potential
(Ps ) for all considered cases. The highest values occur in cases 3, 8 and 10.
On the other hand, the case 3 requires much less material, as it uses only one
fin (the bottom one). Hence, it is simpler and more economical, therefore,
we considered it the best case of the simulations. As we see ahead, further
parameter tests in this paper use just a block and a bottom fin.
Still analysing the results from Table 3, we note that all cases using a
top fin had a soil thermal potential lower than the reference case with no
galvanized material. Physically, it means that a top fin negatively influences
14
Table 3: Annual RMS values of the soil thermal potential for different cases
the results, as it increases the thermal influence of the soil surface in the duct
region. Besides, using only the block, or side fins (cases 1, 4 and 6) do not
affect the soil thermal potential.
It is worth note that the values in Table 3 are annual RMS values, hence
they cannot achieve the peaks of 5◦ C, which occur only in summer or winter,
as in Fig. 5. Indeed, in spring and fall, there are times when the thermal
potential of the soil is close to zero. Moreover, the installation is at a depth
of only 1.6 m, which is not ideal, as we see ahead in this work.
15
(a) Without galvanized material - case 0
16
reducing it by 3 m until reaching an efficiency of less than 70%. Such results
are given in Table 4, and they show how θ reduces by reducing L.
Table 4: Comparison between the duct length (L) and the annual thermal efficiency (θ).
17
Figure 8: Annual comparison of the soil, outlet and air temperatures at different depths.
18
dimensional domain has the same thermal conductivity of the soil. Such
configuration is illustrated in Fig. 9, where the galvanized region has the
shape of a disk of radius r, centered on x0 and z0 (i.e. the duct center).
Here, the galvanized disk area occupies different proportions of the com-
putational domain. The variations took into account the following consider-
ations:
The domain area is the constant Ac =150 m2 , while the disk area (Ad )
varies as a function of r.
The area fraction between disk and domain areas is given by
Ad π r2
ψ= 100% = %.
Ac 1.5
As for the rest of the EAHE modeling, we are adopting the same reference
values from the validation, that is, a duct at depth z0 , with length L0 , and
19
diameter D0 . Basically, the same previous methodology from section 1, but
adopting domains similar to those in Fig. 9 instead of using blocks and fins.
From the results in Table 6, it is clear that the annual RMS values of the
soil thermal potential increase with the reduction of the disk area. The last
line of the same Table also includes the reference case without any galvanized
material. In terms of soil thermal potential, there is no gain in adopting a
disk shape for the galvanized area. On the other hand, as the efficiency goes
from 73% to almost 95% using the galvanized material, there is a gain of
almost 30% in the EAHE thermal potential using just ψ=0.1%.
Another way to see what happens with the introduction of the galvanized
circular area is considering the temperature fields in the computational do-
main. Hence, Fig.10 presents them in the summer (a) and winter (b), taking
ψ=5%. Similarly, Fig.11 shows them for ψ=0.1%.
From Fig.10, one can see that the presence of a large disk (or a large
galvanized region) causes the region around the duct to be strongly influenced
thermally by what happens on the surface. Due to the high conductivity, the
temperatures at the duct position approach those on the surface both, in
summer and winter. Consequently, such unwanted behavior reduces the soil
thermal potential. This situation is well contrasted by observing Fig.11,
where the disk area occupies only 0.1% of the domain. In this case, the
temperatures on the duct region are much different from those on the surface,
thus enabling a higher soil thermal potential.
Figure 12 presents the ambient and outlet air annual temperatures with
the disk occupying 5 and 0.1% of the computational domain. Both cases are
very different throughout the year, as the EAHE thermal potential in the
former is almost half of the latter.
It is also worth taking a look at the result for case 1 in Table 3, where we
used only a square block centered at the duct center. In that case, the block
20
(a) Summer
(b) Winter
area also occupied a small part of the computational domain (about 0.02%).
The soil thermal potential for that case was also equal to the reference case
without any galvanized materials. Such results suggest that using small and
symmetric galvanized areas around the duct center does not increase the
soil thermal potential. However, they are suitable for increasing the EAHE
efficiency and, consequently, the EAHE thermal potential.
In short, increasing the thermal conductivity of the region around the
duct does not imply increasing the soil thermal potential, as it depends on
the temperature difference between the surface and the duct location.
21
(a) Summer
(b) Winter
22
Figure 12: Comparison among annual air and outlet temperatures.
23
H varies between 0 and 5 m, hence SH can range from 0 to 10 m.
The block and the domain have proportional areas. To ensure that, we
assume
2SV
SH = .
3
Thus, there are two limiting cases: (a) SV = 0 m implies SH = 0 m,
and the block converges to a point (the duct center); (b) SV = 15 m
implies SH = 10 m, and the block occupies the entire computational
domain. In short, cases (a) and (b) represent extremes where the entire
domain is composed, respectively, by soil or galvanized material.
The domain area is again the constant Ac =150 m2 , while the block
area is Ab = SH × SV . Hence, the area fraction between block and
domains areas is given by
Ab SH × SV
ψ= 100% = %.
Ac 1.5
24
Figure 14: Thermal potential variations with ψ.
Figure 17 shows the temperature fields in summer and winter for ψ =90%;
while Fig. ?? shows them for ψ =10%. Hence, for ψ =10%, the block helps
to make the region around the duct colder or warmer than the surface. In
such case, the block region improves the connection between the duct and
25
the deeper soil layers. On the other hand, for ψ =90%, the block connects
the regions close to the surface with the deeper ones. In this situation, the
surface temperature significantly influence the process. It can be noted that
the region around the duct has a temperature close to the surface, which
reduces the soil and EAHE thermal potentials.
Such results show once more that only increase the thermal conductivity
on the region around the duct does not imply a better thermal performance.
To improve the soil (and EAHE) thermal potential, it is necessary to think
of suitable shapes for the galvanized area, or galvanized fins coupled to the
ducts.
26
References
[1] R. Hassanzadeh, M. Darvishyadegari, S. Arman, A new idea for improv-
ing the horizontal straight ground source heat exchangers performance,
Sustainable Energy Technologies and Assessments 25 (2018) 138–145.
[6] J. Vaz, Estudo experimental e numérico sobre o uso do solo como reser-
vatório de energia para o aquecimento e resfriamento de ambientes
edificados, Ph.D. thesis, Universidade Federal do Rio Grande do Sul
(UFRGS), Porto Alegre, Brazil (2011).
27
[9] F. Ascione, L. Bellia, F. Minichiello, Earth-to-air heat exchangers for
italian climates, Renewable Energy 36 (2011) 2177–2188.
[10] S. Benkert, F. D. Heidt, D. SchÖler, Calculation tool for earth heat ex-
changers gaea, in: Proceedings Building Simulation, Fifth International
IBPSA Conference 2, -, 1997.
[15] M. N. Özisik, Heat Conduction, John Wiley & Sons, New York, 1993.
[17] T. J. R. Hughes, The finite element method (Linear Static and Dynamic
Finite Element Analysis), Prentice Hall, INC., New Jersey, 1987.
28
[21] L. N. Trefethen, D. Bau, Numerical Linear Algebra, SIAM, 1997.
29