Zonal thermal model of the ventilation of underground transformer

substations: Development and parametric study

Maximiliano Beiza
a
, Juan Carlos Ramos
a,
*
, Alejandro Rivas
a
, Ral Antn
a
,
Gorka S. Larraona
a
, Jon Gastelurrutia
a
, Ivn de Miguel
b
a
TECNUN (University of Navarra), Dept. of Mech. Engineering, Thermal and Fluids Engineering Div., Paseo de Manuel de Lardizbal 13, 20018 San Sebastin,
Spain
b
Ormazabal Distribucin Secundaria, Barrio Basauntz 2, 48140 Igorre, Spain
h i g h l i g h t s
v A zonal model of ventilation of underground transformer substations is presented.
v The development of the zonal model is based on a CFD model.
v A good correlation between the results of the model and the experiments is obtained.
v The model is implemented as a simulation tool for design and optimisation purposes.
a r t i c l e i n f o
Article history:
Received 3 June 2013
Accepted 15 September 2013
Available online 30 September 2013
Keywords:
Thermal model
Zonal model
Transformer substation
Natural ventilation
a b s t r a c t
An algebraic thermal zonal model of the ventilation of underground transformer substations during a
standardised temperature rise test is presented in this paper. The development and adjustment of the
proposed model rely on the analysis of the air ow pattern and temperature distributions obtained by a
more complex model numerically solved by means of CFD techniques. The ow domain of the model
represents a section of the substations divided into several interrelated zones where the mass and the
energy conservation equations are formulated and the generated system of nonlinear algebraic equations
is solved. The model is validated by comparing its results with the ones obtained by the CFD model and
with the experimental results of eight temperature rise tests under different conditions. A parametric
analysis was carried out on the model to prove its utility as an efcient tool to improve and optimise the
thermal performance of transformer substations during the design process. From the parametric study it
has been inferred that the main parameters affecting the ventilation of the substations are the pass area
between the LVeMV zone and the transformer zone, the surface area of the ventilation grilles in the
substation with horizontal ventilation, and the perimeter of the protruding ventilation vents in the
substation with vertical ventilation.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
Underground transformer substations are used to convert
electrical energy from medium to low voltage in electrical power
distribution networks. These buildings are usually made of pre-
fabricated concrete and their main components are the ventilation
grilles, one or two distribution transformers, the Low Voltage (LV)
boards and the Medium Voltage (MV) cubicles. Because of the
power losses occurring in the electrical conversion that takes place
in the transformers and the LV boards, heat is generated. This heat
is dissipated by the air circulating by natural convection through
the substation and by the radiation exchanges with the walls of the
enclosure.
The maximum temperature reached by the top-oil of the
transformer is established by International Standards [1,2] as the
main criterion for evaluating the performance of a transformer
substation. This temperature must be limited in order to extend the
operating life of the transformer. Obtaining a temperature that is
over the limit during the experimental tests run on a real substation
would invalidate the design and new casts would be required. This
is a slow and expensive design procedure that could be avoided by
using a mathematical model of the ventilation in the substation to
perform a simulation to determine the temperatures in the design
stage prior to the experimental tests.
* Corresponding author. Tel.: 34 943 21 98 77; fax: 34 943 31 14 42.
E-mail address: jcramos@tecnun.es (J.C. Ramos).
Contents lists available at ScienceDirect
Applied Thermal Engineering
j ournal homepage: www. el sevi er. com/ l ocat e/ apt hermeng
1359-4311/$ e see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.applthermaleng.2013.09.032
Applied Thermal Engineering 62 (2014) 215e228
One of the rst models of the ventilation of transformer sub-
stations was presented by Menheere in Ref. [3]. It is a very simplied
model that uses one equation for the heat transferred to the venti-
lation air and another equation for the heat dissipated through the
walls of the substation. The main inputs of the model are the power
dissipated by the transformer, the dimensions of the transformer
and the substation, and the heights, surface areas and resistance
coefcients of the inlet and outlet ventilation grilles. The outputs of
the model are estimations of the outowair and of the transformer
mean temperature rises over the ambient temperature.
The next step in the modelling of the ventilation of transformer
substations is the transient equivalent thermal circuit model used
by Radakovic and Maksimovic [4,5] and Iskender and Mamizadeh
[6]. This type of model can be used to obtain the transformer top-oil
temperature and an average or characteristic internal air temper-
ature of experimentally checked substations, but as the parameters
affecting the ventilation are lumped, it is more difcult to perform
an optimisation or a parametric analysis of a transformer substat-
ion in the design stage.
If the objective is to use the model in the design stage of the
substation and with the objective of optimising the ventilation, a
more detailed type of mathematical model must be developed. The
principal effort in this direction is by means of numerically solved
models based on the NaviereStokes and the energy conservation
equations.
In this sense three works stand out in the literature. The rst one
corresponds to Loucaides et al. [7], who use the Finite Element
Method (FEM) to solve the energy and the NaviereStokes equations
in a ow domain corresponding to the air inside a transformer
substation. The model is used to analyse the inuence of the
aperture of the ventilation grills, the transformer load and the
ambient temperature in the air temperature distribution inside the
substation.
The other two works correspond to the research group led by
Ramos et al., who present a preliminary version of their model in
Ref. [8] and the nal version in Ref. [9]. In this last paper they
present a CFD model of the ventilation (air circulation and heat
transfer) of two underground transformer substations solved by
means of the Finite Volume Method (FVM). The model is validated
with the experimental results of eight temperature rise tests car-
ried out under different conditions of ventilation and transformer
power losses. The results of the steady-state simulations serve to
analyse the air ow pattern and the air temperature distributions
inside the substation. Moreover, two important parameters related
to the ventilation performance of the substation that are very
difcult to determine experimentally are obtained from the simu-
lations of the model: a correlation for the air mass ow rate as a
function of the ventilation conditions and the heat dissipated by
the transformer, and correlations for the heat transfer coefcients
on the surfaces of the transformer and the walls of the enclosure.
On the one hand, these CFD models permit design parameters
(size of the substation, dimensions of the walls, and dimensions,
location and types of ventilation grilles, among others) that directly
affect the ventilation of the substation to be taken into account. But
on the other hand, they have huge computational requirements and
consume a great deal of CPU time to perform a single simulation.
These characteristics therefore invalidate the use of these models
for design and optimisation purposes.
Nevertheless, the information obtained in the simulation of the
CFD model can be used to develop an intermediate level model, an
approach known as zonal modelling [10,11], which requires fewer
computational resources and less simulation time, thus allowing its
Nomenclature
Latin letters
A surface area, [m
2
]
c
p
average specic heat, [J/kg K]
C tting coefcient, [e]
e thickness of the wall, [m]
faces number of faces of the zone under consideration, [e]
g gravity acceleration, [9.81 m/s
2
]
h height of the middle point of the zone with respect to
the oor of the substation, [m]
K enclosure class of the substation, [K]
k thermal conductivity, [W/m K]
_ m air mass ow rate, [kg/s]
Nu Nusselt number, [e]
p pressure, [Pa]
Pr Prandtl number, [e]
q heat transfer, [W]
Ra Rayleigh number, [e]
T temperature, [K or

C]
Greek letters
DT temperature rise over ambient temperature, [K]
emissivity, [e]
s StefaneBoltzmann constant, [5.67$10
8
W/m
2
K
4
]
r density, [kg/m
3
]
Subscripts
amb ambient air
base base of the transformer
cond conduction heat transfer
conv convection heat transfer
ns transformer ns
grilles ventilation grilles
hor horizontal face
i zone under consideration
in internal surface
inow inow air
j neighbouring zone
lid lid of the transformer
L characteristic length
max maximum
out external surface
outow outow air
rad radiative heat transfer
source relative to the transformer or to the Low Voltage
boards
transf transformer
upwind relative to the zone where the mass ow rate comes
from
ver vertical face
w each wall of the zone under consideration
wall walls of the substation
Superscripts
r level of iteration on pressure
s level of iteration on temperature
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 216
implementation in a software tool oriented toward designing and
optimising the thermal performance of transformer substations.
This methodology has been successfully applied to the model-
ling of the cooling of distribution transformers by Gastelurrutia
et al. In Ref. [12] they develop a detailed numerical model of the
movement by natural convection of the oil inside distribution
transformers, and based on the oil ow and temperature patterns
obtained, they produced in Ref. [13] an algebraic zonal thermal
model that can be used for design and optimisation purposes.
In the present paper a thermal zonal model of the ventilation of
underground transformer substations is proposed. The develop-
ment of the zonal model is based on the analysis of the air ow
pattern and temperature distributions obtained fromthe numerical
CFD model described in Ref. [9]. The main result provided by the
model is the enclosure class of the substation, that is, the difference
in Kelvin between the transformer top-oil temperature rise over
the ambient temperature inside and outside the substation in
steady-state nominal operating conditions [2]. In order to calculate
the latter parameter, the model interacts with the zonal model of
transformer cooling presented in Ref. [13]. Both models have been
implemented in user-oriented software, which can be used by
transformer substation designers to analyse and optimise the
ventilation and the thermal performance of the substations.
2. Description of underground transformer substations
Fig. 1 shows internal and external views (enumerating the main
components) of the modelled transformer substations identied
henceforth as TS01 (left) and TS02 (right). TS01 has horizontal
ventilation because the grilles are at ground level, and TS02 has
vertical ventilation because the vents protrude from the ground.
The main external dimensions of the transformer substations
are 6.56 m, 2.46 m and 2.79 m in length, width and height,
respectively. Each transformer dissipates 12,200 W in the form of
heat and the LV boards 1400 W.
3. Mathematical model
This section has been subdivided into several subsections to
help the reader understand the development of the zonal model.
In the rst subsection, the owand thermal patterns obtained in
the simulations of the CFD model are analysed. These results have
permitted the ow domain and the division in control volumes of
the zonal model presented in the second subsection to be dened.
In the third subsection the governing equations are presented
with a more in depth description of the mass ow rate equations
and the radiative model, and nally, in the fourth subsection the
boundary conditions are briey summarised.
3.1. Description of the results of the CFD model
Details of the mathematical and numerical characteristics, the
verication and the validation of the developed CFD model can be
found in Ref. [9].
In what follows, the main results of the simulations of the CFD
model that have served to create the algebraic zonal model presented
in the current paper are briey described. The analysis of the air ow
and thermal patterns of the simulations of the model is presented
because it has been used to dene the ow domain and the control
volumes of the zonal model. Figs. 2 and 3 represent the temperature
contours (left) and the velocity vectors (right) for the simulations of
tests TS01-E02 and TS02-E06, respectively. In both tests the ventila-
tion condition is the standard; that is, all the grilles are installed and
the transformer power losses are 100% of the nominal value.
In analysing the temperature and velocity elds shown in Figs. 2
and 3 together, it can be said that the fresh air entering the sub-
stations through the inow grilles in the LVeMV zone, (A), goes
directly to the bottom for two reasons. First, fresh air has higher
density and it weighs more than hot air. Second, the driving force
behind the air owis the heat dissipated by the transformer, and so
the air passes to the transformer zone through the free space below
the protective plate that separates the LVeMV zone from the
transformer zone, (B). After passing between the ns of the trans-
former, (C), the cooling air is redirected to the outowwindows and
from there enters the external vents and leaves the substations
through the grilles, (D).
It can be also observed that in the top part of the LVeMV zone of
the substations there is warm air at an intermediate temperature
(about 35e37

C), which is not participating in the ventilation of
the substation (low velocities in Figs. 2 and 3 (right)). This air is
heated by the power dissipated by the LV boards. In the transformer
zone and above it, the air is at a temperature of around 50

C, and
the plumes of hot air (z60

C) coming from the ns of the trans-
former and going to the outow windows can be observed.
Finally, and as pointed out in Refs. [9], the most important re-
sults of the simulations of the model are those that can barely be
determined experimentally but are intrinsically related to the
ventilation and thermal performance of the substations and that
will be used in the development of the zonal model: the mass ow
rate of the air circulating through the substation and the heat
transfer coefcients over the surfaces of the transformer and the
walls of the enclosure.
Fig. 1. Conguration of the two modelled transformer substations with their main components: 2 distribution transformers (1), the Low Voltage boards (2), the Medium Voltage
switchgears (3), the protective metallic plates (4), the concrete walls of the enclosure (5), the prefabricated concrete vents (6), the inow (7) and outow (8) windows and louver
grilles in the substation with horizontal ventilation (TS01), and the inow (9) and outow (10) horizontal (left image) or vertical (right image) external ventilation grilles.
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 217
The eight experimental temperature rise tests carried out to
validate the CFD model were done under different ventilation
conditions (removing the grilles or closing the windows) and
transformer power losses (100% or 50%) in order to capture the
inuence of these input parameters on the ventilation and the
thermal performance of the substations. From the analysis of the
results of the eight simulations of the experimental tests, it was
concluded that the previous parameters could be correlated with
the quantity of air owing through the substations through Eq. (1).
_ m = 0:0358$q
0:297
conv;transf
$

C
d;grilles
$A
grilles

0:700
total
(1)
In a similar process, empirical correlations from the literature
for the average Nusselt number on the lid and on the base of the
transformer and on the walls of the enclosure were tted with the
results of the simulations. However, for the ns of the transformer
an average correlation for the local Nusselt number in the vertical
direction was tted [9].
3.2. Flow domain and control volumes of the zonal model
After analysing the air velocity elds obtained in the eight sim-
ulations of the CFD model (the results for two simulations in two
perpendicular planes are shown in Figs. 2 and 3 (right)) it was
concluded that although the real ow domain was three-
dimensional, the main air ow pattern was two-dimensional. This
conclusion is perfectly applicable for substationTS02, where the inlet
and outlet protruding vents are in the same vertical plane, and it can
be also applied to substation TS01 if the plane where the pre-
fabricated concrete inlet vent lies is folded down by 90

, as shown in
Fig. 4 (left). Therefore, the ow domain of the zonal model is the air
inside a vertical plane cutting the substation longitudinally and
including the inlet and outlet vents. Moreover, only one half of the
substation is included in the ow domain because, as with the CFD
model, geometrical and thermal symmetry with respect to a vertical
plane that cuts the substation width-wise can be assumed.
The 39 control volumes or zones in which the ow domain of
both substations has been divided are shown in Fig. 4. In Fig. 4 (left)
the zones are superimposed over the velocity vectors coloured by
velocity magnitude in the two perpendicular vertical planes cutting
the inlet and outlet vents of substation TS01 shown in Fig. 2. The
plane of the inlet vents, titled front view and containing zones 1
to 20, has been fold down by 90

, as explained in the previous

paragraph, to coincide with the plane of the outlet vents, titled
lateral view and containing zones 21 to 39. Fig. 4 (right) shows the
zones superimposed over the vertical plane of substation TS02 that
contains the inlet and outlet protruding vents.
Fig. 2. Contours of temperature in

C (left) and velocity vectors coloured by velocity magnitude in m/s (right) at two perpendicular planes obtained in the simulation of the heating
test TS01-E02.
Fig. 3. Contours of temperature in

C (left) and velocity vectors coloured by velocity magnitude in m/s (right) at two perpendicular planes obtained in the simulation of the heating
test TS02-E06.
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 218
Although the zonal model is 2D for the air ow, it is 3D for the
heat transfer. That is, each zone represented in Fig. 4 is not a rect-
angle, but a rectangular prism where the dimension perpendicular
to the plane of the gure is the width of the substation. In this way
the zonal model accounts for the heat transfer through the lateral
walls of the substation.
The number of zones that the owdomain has been divided into
has been dened by taking into account the fact that the zonal
model must be capable of considering the range of variation in the
dimensions of the substation and its components and of repre-
senting the evolution of the air mass ow-rates and temperature, as
obtained by the CFD model and shown in Figs. 2 and 3.
Fig. 4 helps explain the ow domain division. For example, in
the LVeMV zone of the substation, vertically dividing the zone
into 3 columns was established as the minimum needed to take
into account the development air ow inside the zone. In this
sense, the rightmost column of control volumes (in Fig. 4) cor-
responds to the ow entering the substation through the inlet
vent, the central column corresponds to the ow evolution in the
LVeMV zone, and the leftmost column corresponds to the ow
passing into the transformer zone. The central column could
have been divided into more control volumes, but the small
temperature gradient obtained by the CFD model (Figs. 2 and 3
left) made it unnecessary. The horizontal division of 6 rows of
control volumes in the LVeMV zone can be explained in the
same way: the top row corresponds to the entering ow, the
second, third and fourth rows correspond to the development of
the ow and to the height of the LV board, the fth row corre-
sponds to the air ow passing into the transformer zone and the
sixth row corresponds to the lowest part of the substation.
The transformer zone of the substation is vertically divided into
three columns in order to include in the zonal model the thermal
plumes generated from the lateral ns and the lid of the trans-
former. Again, the horizontal division of 6 rows of control volumes
can be explained as follows: the top rowcorresponds to the top part
of the substation, the second to the ow leaving the substation, the
third and fourth rows to the air owevolution over the transformer,
the fth row corresponds to the ns and the bottom row to the
bottom part of the transformer.
The dimensions of the 39 zones or control volumes vary when
the dimensions of the substations and their components are varied
by the designer, but the number of zones is kept constant in each
simulation of the model.
Since the validation of the zonal model was acceptable (their
results are compared with the experimental results in Section 6), it
was not considered necessary to include more control volumes in
the zonal model.
3.3. Governing equations
The mass and energy conservation equations in their steady-
state formulation, Eqs. (2) and (3), are applied to the control vol-
umes into which the ow domain has been divided, with the
objective of calculating the pressure and temperature in each zone
and the mass ow rates between them.
X
faces;i
j =1
_ m
ji
= 0 (2)
X
faces;i
j =1

_ m
ji
$c
p
$T
upwind
q
conv;wall;j
q
conv;source;j

= 0 (3)
Eq. (2) says that the sum of the air mass ow rates entering or
leavingonezonethroughthefaces that areincontact withother zones
equals zero. Eq. (3) establishes that the sumof the net air owthermal
energy rates entering or leaving one zone through its boundaries that
are incontact with other zones, plus the sumof the heat interchanged
by convection with the internal surfaces of the external walls of the
substations, plus the convection heat from the walls of the trans-
former or of the LV boards, the heat sources, must be zero.
In the previous equations the subscripts i and j represent the
zone considered and the neighbouring zones, respectively; the
mass owrates, _ m
ji
, are positive if they enter the zone and negative
if they leave it; c
p
is the specic heat of air evaluated at the average
temperature of the whole domain; and the temperature T
upwind
corresponds to the zone from which the air mass ow rate comes.
Additionally, the mass and the energy conservation equations
and the internal radiation exchange balance equation must be
fullled for the whole ow domain, as represented by Eqs. (4)e(6).
_ m
inflow
_ m
outflow
= 0 (4)
q
total;sources
_ m
inflow
$c
p
$

T
outflow
T
inflow

q
total;out
= 0
(5)
X
all faces
w=1
q
rad;in;w
= 0 (6)
Fig. 4. Denition of the control volumes of the zonal model of TS01 (left) and TS02 (right) superimposed over the velocity elds in a plane cutting the inlet and outlet vents.
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 219
In Eq. (5) q
total,sources
represents the total heat, by convection and
radiation, dissipated by the transformer and the LV boards and
q
total,out
the heat released by convection and radiation from the
external surfaces of the walls of the substation. Eq. (6) represents
the fact that the sumof the radiation heat transfer of all the faces of
the zones of the model in contact with the walls of the enclosure
must be null.
3.3.1. Mass ow rate equations
The air mass ow rates through the faces of the zones where
there is no thermal source, the so-called pressure cells, are calcu-
lated proportional to the pressure difference between the zones
sharing the face. For the horizontal mass ow rates through the
vertical faces Eq. (7) is used, and for the vertical ones across the
horizontal faces Eq. (8) is employed, which includes the hydrostatic
pressure difference between cells.
_ m
ji;hor
= 0:83$r
upwind
$A
ji
$

p
j
p
i

0:5
(7)
_ m
ji;ver
= 0:83$r
upwind
$A
ji
$
h
p
j
p
i
0:5$g$

r
j
$h
j
r
i
$h
i
i
0:5
(8)
In the previous equations, the values of the discharge coef-
cient, 0.83 m/s Pa
0.5
, and the power coefcient of the pressure
difference, 0.5, have been selected according to [11]. The density in
each zone is calculated as a function of the pressure and the tem-
perature through the Ideal Gas Law.
To know what temperature and density should be considered in
Eqs (3) and (7), it is necessary to determine the upwind zone, which
is where the ow originates. This is done by means of the calcu-
lation of the sign of the pressure difference between adjoining
cells:
if sign

p
j
p
i

= 10T
upwind
= T
j
and r
upwind
= r
j
; if sign

p
j
p
i

= 10T
upwind
= T
i
and r
upwind
= r
i
:
In the case of the two cells in contact with the transformer in
Fig. 4, the mass ow rate leaving them is calculated with Eq. (1). As
pointed out in the previous section, this equation has been tted
with the results of the simulations of the CFD model. The expres-
sion has been deduced from the consulted literature [11], where it
is usually assumed that the relationship between the air mass ow
rate in a thermal plume and the heat source is a power of 1/3. But in
the transformer substations, there are two other parameters that
affect the ventilation air ow, the surface area and the discharge
coefcient of the grilles. For this reason they have been introduced
into Eq. (1), as explained in Ref. [9].
3.3.2. Radiative heat transfer modelling
In the energy conservation equation of the whole domain, Eq.
(5), the external radiation heat transfer is taken into account. This is
modelled by assuming that the external surfaces of the walls of the
substation are grey-diffuse with a constant value of the emissivity
and exchange radiation with a black body at the temperature of the
ambient air, Eq. (9).
q
rad;out;w
= s$
out;w
$A
w
$

T
4
amb
T
4
wall;out;w

(9)
The consideration of the external radiation heat transfer in the
zonal model implies that a newrestriction equation must be added
for each external solid face of the zones that the ow domain has
been divided into. This restriction equation establishes that the
sum of the external convection and radiation heat transfers equals
the one-dimensional heat conduction through the wall of the
substation. The restriction is presented in Eq. (10).
q
cond;w
= k
wall;w
$A
w

T
wall;in;w
T
wall;out;w

e
wall;w
=

q
conv;out;w
q
rad;out;w

(10)
Taking into account the temperatures reached on the surface of
the transformer (around 80

C) the internal radiation heat ex-
changes must be also included in the zonal model. These have been
done using a modication of the radiative model proposed by
Walton in Ref. [15]. This model is used to calculate in a simplied
manner the radiative interchanges between the walls of any type of
enclosure, assuming that each surface that the enclosure has been
divided into exchanges radiation with a ctitious surface whose
radiation magnitudes (surface area, emissivity and temperature)
are weighted averages of all the surfaces of the enclosure.
Waltons radiation model, and its implementation as described
in Refs. [10], does not take into account the fact that the ctitious
surface should not include the surfaces of the enclosure not viewed
by the surface whose radiative heat transfer is been calculated. That
is, the radiation view factors are not considered. In the present
implementation, the fact that there are surfaces that cannot ex-
change radiation with others because they do not see each other
and, therefore, they cannot be included in the ctitious surface has
been taken into account. This condition has been introduced by
means of simplied view factors that have a value of 1 or 0,
depending on whether the surfaces see each other (1) or not (0).
Because the internal radiative model works with ctitious sur-
faces, an imbalance is introduced in the radiation exchange balance
equation, Eq. (6). To correct it, this imbalance is shared out among
all the solid faces of the zones of the model in an area-weighted
way, as proposed by Walton [15]. The correction of the internal
radiation imbalance introduces an iterative process in the resolu-
tion of the model that is stopped when its value reaches a
minimum.
In the same way as for the external radiation, the inclusion of the
internal radiation heat transfer in the zonal model implies that a
new restriction equation must be added for each internal solid face
of the zones of the model. This restriction equation establishes that
the sum of the internal convection and radiation heat transfers
equals the one-dimensional heat conduction through the wall of
the substation. The restriction is presented in Eq. (11).
q
cond;w
= k
wall;w
$A
w

T
wall;in;w
T
wall;out;w

e
wall;w
=

q
conv;in;w
q
rad;in;corrected;w

(11)
3.4. Boundary conditions
In this section the different types of boundary conditions
imposed on the limits of the ow domain of the zonal model are
summarised.
On the external surfaces of the walls of the substation mixed
conditions of convection and radiation are imposed; the exception
is the external surfaces corresponding to the oor, where the
ambient air temperature is xed. The external radiation transfer is
modelled as explained in Section 3.3.2. For the convection heat
transfer, the average heat transfer coefcients are calculated by
using empirical correlations from the literature [14], tted with the
results of the eight temperature rise tests simulated with the CFD
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 220
model. The correlations are shown in Eqs. (12) and (13) for the
vertical and horizontal walls, respectively.
Nu
ver;out;wall
= C$
0
B
@0:825
0:387$Ra
1=6
L

0:492
Pr

9=16

8=27
1
C
A (12)
Nu
hor;out;wall
= C$0:54$Ra
1
4
L

10
4
_ Ra
L
_ 10
7

Nu
hor;out;wall
= C$0:15$Ra
1
3
L

10
7
_ Ra
L
_ 10
11

(13)
These heat transfer coefcients are non-linear and vary
throughout the resolution process of the model because the
thermo-physical properties of the air are calculated at the lm
temperature, the average temperature of the ambient air and the
surface of the wall, the latter being a result of the simulation.
For the internal surfaces of the walls, mixed conditions of con-
vection and radiation are again used. They are assumed to be grey-
diffuse with constant emissivity and the radiation exchange is
calculated with a modication of Waltons model, as explained in
Section 3.3.2. For the convective heat exchange, the different
average heat transfer coefcients have been obtained by using
empirical correlations from the literature [14] for the walls of the
enclosure and the LV boards and using the correlation presented in
Ref. [9] for the surface of the ns of the transformer. These corre-
lations are shown in Eqs. (12)e(14).
Nu
hor;in;wall
= C$0:27$Ra
1
4
L

10
5
_ Ra
L
_ 10
10

(14)
In all the cases the correlations have been tted with the results
of the eight temperature rise tests simulated with the CFD model.
These heat transfer coefcients are non-linear and vary throughout
the resolution process of the model because the thermo-physical
properties of the air are calculated at the lm temperature, the
average temperature of the air and the surface of the wall, both
being results of the model.
On the faces of the cells in contact with the LV boards, the power
dissipated by them is imposed proportionally to the surface area of
each face. The faces of the zones in contact with the protective
metallic plate that separates the transformer from the LVeMV
zones are assumed to be adiabatic. On both types of faces, the ra-
diation and convection boundary conditions are of the same type as
for the internal surface of the walls of the substation described in
the previous paragraph.
On the faces of the cells in contact with the three parts that the
transformer is divided into (lid, ns and base), the boundary con-
dition imposed is the proportional part of the power dissipated by
the transformer. The distribution of the power dissipated by the
transformer between the three parts it is divided into is a result of
the zonal model of the transformer described in Refs. [13], because
the zonal model of the substation interacts with the zonal model of
the transformer, as will be described later. Again, on these types of
faces the radiation and convection boundary conditions are of the
same type as for the rest of the solid surfaces of the substation.
Finally, on the faces of the cells corresponding to a substation
ventilation grille, its discharge coefcient is used to calculate the air
mass ow rate through each face with Eq. (7) or Eq. (8), depending
on whether they are vertical or horizontal.
4. Implementation of the model and resolution procedure
The thermal zonal model that was described in the previous
section was implemented in a software application to check its real
performance. The main input to the model consists of the ambient
temperature, the most relevant geometric characteristics of the
enclosure, the transformer and the ventilation grilles, and the po-
wer losses of the transformer. The output of the model consists of
the substation enclosure class, the ventilation air mass ow rate,
the air temperature at the outlet grille and the maximum air
temperature inside the substation, the average temperatures of the
different walls of the substation, and the distribution of the heat
dissipated by the ventilation air and through the walls of the
enclosure.
A Microsoft Excel

spreadsheet was selected, along with the

built-in Premium Solver, to implement the mathematical model.
This widely used programme offers a user-friendly environment
and includes a lot of programming options through use of Visual
Basic for Applications.
To use the Premium Solver, all the previous equations are reor-
dered and the system is expressed as an optimisation problemwith
several restrictions. The reformulation of the problem is expressed
by Eqs. (15)e(18) in a general form.
min
x
f (x) : (15)
h
b
(x) = 0; b = 1; 2; .B (16)
g
d
(x) _ 0; d = 1; 2; .D (17)
x
U
i
_ x
i
_ x
L
i
; i = 1; 2; .N (18)
To obtain the pressures and the mass ow rates, the function to
be minimised drepresented by f(x) dis the sum of the squares of
the mass imbalance of each zone. The restrictions h
b
(x) and g
d
(x)
are given by Eq. (2), and there is no limit to the pressure values. In
this system, 38 pressures and 75 mass ow rates are solved.
To obtain the temperatures the function to be minimised is the
total energy imbalance represented by Eq. (5), where the re-
strictions are given by Eqs. (3), (10) and (11), and there is no limit to
the values of the variable. The resolution of this systemof equations
results in 173 temperatures, 39 of the air zones and 134 of the in-
ternal and external surfaces of the walls of the enclosure.
An enhanced version of the Generalized Reduced Gradient algo-
rithm (GRG2) [16] was considered to be the best choice among all
the resolution engines that are available in the Premium Solver to
deal with this problem. The initial values of the independent var-
iables are calculated using quadratic extrapolation, the derivatives
are evaluated by central differentiation, and the searching direction
is found by means of the Conjugated Gradient Method.
The equations in the developed algebraic equation system are
coupled and have several nonlinearities that cause an iterative
resolution procedure to be implemented in which the mass con-
servation, the energy equation and the internal radiation balance
equation are sequentially solved, as is shown in the owchart in
Fig. 5.
Firstly, to uncouple the mass and the energy conservation
equations during the iterative resolution procedure, the air density
of every zone is kept constant and is not recalculated until the new
temperatures from the solution of the energy conservation equa-
tion are obtained.
The resolution of the mass conservation equation is done in an
iterative way, because the term for pressure difference in Eqs. (7)
and (8) is linearized to avoid numerical problems in the case that
a negative value to the power of a number less than 1 is obtained.
The way the equation is linearized is shown in Eq. (19).
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 221

p
r
j
p
r
i

0:5
=

p
r
j
p
r
i

1
$

p
r1
j
p
r1
i

0:5
(19)
where the superscripts r and r 1 represent the actual and the old
level of iteration.
Inside every iterative level the mass conservation equation is
solved by checking that the sum of the squares of the mass
imbalance of all the zones is zero. The iterative procedure of the
calculation of the mass ow rates is halted when the difference
between two consecutive iteration levels of the nonlinear term of
the pressure difference is less than a prescribed value
(0.0001 Pa
0.5
).
Then, the energy conservation equation is solved by making null
the difference of the power dissipated by the transformer and the
LV boards with the heat dissipated by the ventilation air and
through the walls of the substation.
The rst time this equation is solved, an imbalance appears in
the internal radiation exchange balance equation, as explained in
Section 3.3.2. This imbalance is distributed among all the faces of
the zones, and this makes it necessary to recalculate the temper-
atures with the energy conservation equation corrected with this
imbalance. The procedure is repeated several times until the dif-
ference of the radiation imbalances between two consecutive it-
erations is less than a minimum value (1 W).
Finally, once the new temperatures have been obtained, it is
veried that the difference between the new and the old value of
the temperature of the cell representing the outow air of the
substation is less than a minimum value (0.5 K). In the negative
case, the densities are recalculated with the new temperatures and
the entire calculation procedure is repeated.
The above calculation procedure is represented in the owchart
in Fig. 5.
5. Interaction between the zonal models of the transformer
and of the substation
In order to determine the enclosure class of the substation, the
thermal zonal model of the transformer substation described in
this paper has to interact with the zonal model of the transformer
described in Ref. [13]. This latter model has two calculation modes:
outside mode, where the transformer is outside the substation, and
inside mode, where it is inside.
The substation zonal model interacts with both calculation re-
gimes following the procedure represented in Fig. 6.
Fig. 5. Flowchart of the iterative resolution procedure of the zonal model.
Fig. 6. Flowchart of the interaction between the zonal models of the transformer and
the substation.
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 222
The simulation of one temperature rise test of one substation
with the zonal model, including the interaction with the zonal
model of the transformer, takes around 20 min on a PC with a
processor running at 3.10 GHz and with 4 GB of RAM memory. The
same simulation with the CFD model can run between 4 and 6
weeks in a calculation server with 4 double-core processors at
2.80 GHz and 64 GB of RAM memory.
6. Validation of the model
The model described in the previous sections has been run for
the eight temperature rise tests described in Ref. [9]. The validation
has been carried out by comparing the results of the zonal model
with the ones obtained in the experimental tests. Moreover, the
results of the zonal model have been also compared with the re-
sults of the simulations of the CFD model. In Figs. 7 and 8 a quali-
tative comparison of the temperatures obtained by the CFD and the
zonal models in the simulations of two tests, one for each sub-
station, is presented. It can be observed a good correlation between
both models.
The validation and comparison of the numerical results is pre-
sented in Tables 1 and 2. From the results shown in them it can be
concluded that the validation of the model is good. For the venti-
lation air mass ow rate, the maximum difference between the
zonal model and the CFDmodel occurs for test TS01-E02 and is 10%.
For the enclosure class of the substation the results are very good
for three of the four tests in which this parameter can be calculated,
with a difference around 1 K. For the outow air temperature rise,
DT
outow
, the comparison with the CFD model is good in general
(differences around or less than 2 K in ve tests), and it is inac-
ceptable in test TS01-E04 (a difference of 10 K). As pointed out in
Refs. [9], the reason could be related to the presence of a wide re-
gion of reverse ow in the outlet grilles obtained in the simulation
of the CFD model that the zonal model cannot capture and that was
smaller in the experimental test than in the simulation.
For the maximum air temperature rise inside the substation,
DT
max
, the results of the zonal model are acceptable, if the fact that
we are comparing temperatures measured at a particular point in
the case of the CFD model and the experiments with the temper-
ature of a zone in the zonal model is taken into account. For test
TS02-E06 the difference between the zonal and the CFD model is
7 K, but it seems that can be imputed to the latter model, as the
difference between the zonal model and the experiment is very
small at 0.4 K.
The results for the average temperature rises on the surfaces of
the lid, ns and base of the transformer can also be considered
acceptable, as no discordant values have been found. For the
average temperature rise of the internal surface of the walls of the
Fig. 7. Comparison of the CFD and the zonal model temperature results (in

C) for the simulation of test TS01-E02.
Fig. 8. Comparison of the CFD and the zonal model temperature results (in

C) for the simulation of test TS02-E06.
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 223
Table 1
Comparison of the zonal model results with the experimental and the CFD model results for the simulations of the temperature rise tests for substation TS01.
TS01-E01 TS01-E02 TS01-E03 TS01-E04
Zonal CFD Exp. Z C Z E Zonal CFD Exp. Z C Z E Zonal CFD Exp. Z C Z E Zonal CFD Exp. Z C Z E
_ m
air
[kg/s] 0.4013 0.3863 e 4% e 0.3474 0.3157 e 10% e 0.2816 0.2623 e 7% e 0.1557 0.1498 e 4% e
K [K] 13.1 e 9.1 e 4.0 16.1 e 17.1 e 1.0 10.6 e e e e 18.9 e e e e
DT
outow
[K] 30.1 30.0 30.7 0.1 0.6 34.2 36.0 39.1 1.8 4.9 20.9 21.7 15.8 0.8 5.1 33.7 23.7 29.8 10.0 3.9
DT
max
[K] 30.9 29.9 31.9 1.0 1.0 35.5 36.7 38.0 1.2 2.5 21.9 23.1 24.4 1.2 2.5 38.2 35.2 34.5 3.0 3.7
DT
lid
[K] 67.1 e 65.1 e 2.0 70.1 e 72.7 e 2.6 41.8 e 45.1 e 3.3 50.3 e 52.5 e 2.2
DT
ns
[K] 63.5 e 59.1 e 4.4 66.8 e 36.7 e 0.1 39.4 e 40.4 e 1.0 48.9 e 48.9 e 0.0
DT
base
[K] 28.0 e 26.1 e 1.9 29.5 e 32.5 e 3.0 17.5 e 20.2 e 2.7 22.0 e 26.9 e 4.9
DT
wall
[K] 19.8 17.1 17.3 2.7 2.5 21.9 23.4 22.1 1.5 0.2 12.7 13.3 13.2 0.6 0.5 19.62 18.8 18.8 0.8 0.8
_ m
air
: air mass ow rate; K: enclosure class of the substation; DT
outow
: outow air temperature rise over the ambient temperature; DT
max
: maximum air temperature rise inside the substation over the lid of the transformer;
DT
lid
: maximumtemperature rise of the transformer lid; DT
ns
: average temperature rise of the transformer ns; DT
base
: maximumtemperature rise of the transformer base; DT
wall
: average temperature rise of the internal lateral
walls of the enclosure; Z C: zonal minus CFD results; Z E: zonal minus experimental results.
Table 2
Comparison of the zonal model results with the experimental and the CFD model results for the simulations of the temperature rise tests for substation TS02.
TS02-E05 TS02-E06 TS02-E07 TS02-E08
Zonal CFD Exp. Z C Z E Zonal CFD Exp. Z C Z E Zonal CFD Exp. Z C Z E Zonal CFD Exp. Z C Z E
_ m
air
[kg/s] 0.3430 0.3646 e 6% e 0.2995 0.3047 e 2% e 0.2433 0.2501 e 3% e 0.1934 0.1778 e 9% e
K [K] 12.2 e 11.1 e 1.1 17.4 e 16.9 e 0.5 11.5 e e e e 17.4 e e e e
DT
outow
[K] 35.3 33.9 30.7 1.4 4.6 39.2 35.0 40.4 4.2 1.2 23.8 20.7 25.3 3.1 1.5 28.7 31.0 32.4 2.3 3.7
DT
max
[K] 36.2 32.1 33.4 4.1 2.8 40.9 33.9 40.5 7.0 0.4 25.1 23.5 25.5 1.6 0.4 30.8 31.4 31.6 0.6 0.8
DT
lid
[K] 66.6 e 66.4 e 0.2 71.7 e 73.4 e 1.7 42.8 e 45.6 e 2.8 48.5 e 51.3 e 2.8
DT
ns
[K] 62.3 e 64.1 e 1.8 68.4 e 71.1 e 2.7 40.5 e 43.6 e 3.1 47.4 e 49.5 e 2.1
DT
base
[K] 27.6 e 26.9 e 0.7 30.3 e 32.5 e 2.2 18.0 e 20.4 e 2.4 21.0 e 25.9 e 4.9
DT
wall
[K] 20.4 18.0 18.3 2.4 2.1 23.0 23.4 23.3 0.4 0.3 13.1 13.7 14.1 0.6 1.0 16.0 18.7 18.5 2.7 2.5
_ m
air
: air mass ow rate; K: enclosure class of the substation; DT
outow
: outow air temperature rise over the ambient temperature; DT
max
: maximum air temperature rise inside the substation over the lid of the transformer;
DT
lid
: maximumtemperature rise of the transformer lid; DT
ns
: average temperature rise of the transformer ns; DT
base
: maximumtemperature rise of the transformer base; DT
wall
: average temperature rise of the internal lateral
walls of the enclosure; Z C: zonal minus CFD results; Z E: zonal minus experimental results.
M
.
B
e
i
z
a
e
t
a
l
.
/
A
p
p
l
i
e
d
T
h
e
r
m
a
l
E
n
g
i
n
e
e
r
i
n
g
6
2
(
2
0
1
4
)
2
1
5
e
2
2
8
2
2
4
enclosure, the results are good, with differences that are less than
2.5 K in all the cases.
7. Results and discussion
In order to show the potential and utility of the proposed zonal
thermal model for design purposes, a broad parametric study was
carried out, taking as reference for both substations the nominal
operating conditions, 100% of transformer power losses and standard
ventilation grilles, with an ambient temperature of 20

C (the bars
labelled H00 in Fig. 9 and V00 in Fig. 10 below). Fifteen parameters
were changed for substationTS01, resulting in 32 simulations, and 11
parameters for TS02, resulting in25 simulations. InTables 3 and 4 the
correspondence between the number of the simulation, the param-
eters changed and their range of variation is presented. All the pa-
rameters were modied in a realistic range and taking into account
the dimensions of the real components. The output result of the
model chosen to gauge the inuence of the variations was the
enclosure class of the substation. In the Introduction this parameter
was theoretically dened and a numerical example is nowpresented
to better illustrate its physical meaning. Prior to the 8 experimental
tests presented in Tables 1 and 2, the transformer was tested outside
Fig. 9. Enclosure class for several simulations of the zonal model of substation TS01.
Fig. 10. Enclosure class for several simulations of the zonal model of substation TS02.
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 225
the substations. During this test the ambient temperature was 14.6

C
and the temperature of the top-oil was 70.7

C. This means that the
transformer top-oil temperature rise over ambient temperature was
70.7 14.6 = 56.1 K. During test TS01-E01 in Table 1, the ambient
temperature was 16.1

C and the top-oil temperature of the trans-
former inside the substation was 81.3

C, which is a temperature rise

of 81.3 16.1 =65.2 K. The enclosure class of the substation can be
calculated by subtracting the top-oil temperature rises inside and
outside the substation. In this case, the enclosure class, K, of sub-
station TS01 under test E01 was 65.2 56.1 =9.1 K. In other words,
the substation is hampering the cooling of the transformer by
generating an increase of 9.1 K in the top-oil temperature rise with
respect to testing the transformer outside the substation.
The parameters varied that affected by less than 1 K the value of
the enclosure class of both substations were: the increase or
decrease of the ambient and oor temperatures; the increase or
decrease of the thickness of the lateral walls and ceiling; the in-
crease or decrease of the thermal conductivity of the material of the
walls; the increase or decrease of the total dimensions (length,
width and height) of the substation; the increase of the length of
the space for the transformer; and the decrease of the head loss
coefcient of the inlet and outlet expanded metal grilles (the grilles
with the higher coefcient).
The parameters most inuencing the output variables of the
model are shown in Figs. 9 and 10 for substations TS01 and TS02,
respectively, and are discussed in the following paragraphs.
It can be observed in Fig. 9 that in the reference case, simulation
H00, the K value of the substation is 15.5 K. Simulation H20 cor-
responds to the maximum physically possible reduction (77% with
respect to the original value) of the pass area between the LVeMV
zone and the transformer zone of the substation. The K value in-
creases by 7.5 K, denoting the importance of this parameter. In the
modelled substations this area cannot be increased due to
geometrical restrictions, making it impossible to gauge the gener-
ated reduction of the K value.
In simulation H26, the head loss coefcient of both inlet and
outlet expanded metal grilles has been doubled and the increase in
K is only by 1.8 K, indicating that this is not a crucial parameter.
In simulations H27 and H28, the surface area of the inlet and
outlet windows, respectively, was reduced by half and in simulation
H29, both were reduced by half simultaneously. The variation in the
K value is greater when the inlet windows are reduced because
their number, 6, is smaller than the number of outlet windows, 8.
Simulations H30, H31 and H32 correspond to a 50% increase in the
surface area of inlet, outlet and both ventilation windows, respec-
tively. Again the greater reduction in the K value is obtained when
the inlet windows are varied. From simulations H27 to H32, it can
be also concluded that when both inlet and outlet surface areas of
the ventilation windows are modied, the variation in the K of the
substation is more or less the sum of the increases or the decreases
obtained with each modication separately.
Fig. 10 shows the results of the parametric analysis for sub-
station TS02. In this case, in the simulation corresponding to a
reduction of the pass area between the LVeMV zone and the
Table 3
Parameters changed and range of variation in the simulations of the parametric
study carried out for substation TS01.
Simulation Parameters changed Range of variation
H01 T
amb
and T
oor
0

C and 0

C
H02 T
amb
and T
oor
40

C and 40

C
H03 T
amb
and T
oor
20

C and 15

C
H04 T
amb
and T
oor
20

C and 25

C
H05 p
amb
7%
H06 p
amb
12%
H07 Thickness of the lateral wall 113%
H08 Thickness of the lateral wall 200%
H09 Thickness of the ceiling 42%
H10 Thickness of the ceiling 31%
H11 Wall thermal conductivity 50%
H12 Wall thermal conductivity 50%
H13 Substation height 28%
H14 Substation height 12%
H15 Substation width 35%
H16 Substation width 7%
H17 Substation length 42%
H18 Substation length 11%
H19 Transformer zone length 11%
H20 Height of the pass between
LVeMV and transformer zone
77%
H21 Head loss coefcient of inlet
expanded metal grille
50%
H22 Head loss coefcient of outlet
expanded metal grille
50%
H23 Head loss coefcient of both
inlet and outlet expanded
metal grilles
50%
H24 Head loss coefcient of inlet
expanded metal grille
100%
H25 Head loss coefcient of outlet
expanded metal grille
100%
H26 Head loss coefcient of both
inlet and outlet expanded metal grilles
100%
H27 Surface area of inlet windows 50%
H28 Surface area of outlet windows 50%
H29 Surface area of inlet and outlet windows 50%
H30 Surface area of inlet windows 50%
H31 Surface area of outlet windows 50%
H32 Surface area of inlet and outlet windows 50%
Table 4
Parameters changed and range of variation in the simulations of the parametric
study carried out for substation TS02.
Simulation Parameters changed Range of variation
V01 T
amb
and T
oor
0

C and 0

C
V02 T
amb
and T
oor
40

C and 40

C
V03 T
amb
and T
oor
20

C and 15

C
V04 T
amb
and T
oor
20

C and 25

C
V08 Thickness of the lateral wall 200%
V15 Substation width 35%
V20 Height of the pass between
LVeMV and transformer zone
77%
V21 Head loss coefcient of inlet
expanded metal grille
50%
V22 Head loss coefcient of outlet
expanded metal grille
50%
V23 Head loss coefcient of both
inlet and outlet expanded metal grilles
50%
V24 Head loss coefcient of inlet
expanded metal grille
100%
V25 Head loss coefcient of outlet
expanded metal grille
100%
V26 Head loss coefcient of both
inlet and outlet expanded metal grilles
100%
V27 Surface area of inlet vents 15%
V28 Surface area of outlet vents 15%
V29 Surface area of inlet and outlet vents 15%
V30 Surface area of inlet vents 45%
V31 Surface area of outlet vents 45%
V32 Surface area of inlet and outlet vents 45%
V33 Surface area of inlet vents and
head loss coefcient of inlet
expanded metal grille
50% and 35%
V34 Surface area of outlet vents and
head loss coefcient of outlet
expanded metal grille
50% and 10%
V35 Surface area of inlet and outlet
vents and head loss coefcient of
inlet and outlet expanded metal grilles
50%, 50%,
35% and 10%
V36 Surface area of inlet vents 50%
V37 Surface area of outlet vents 50%
V38 Surface area of inlet and outlet vents 50%
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 226
transformer zone of the substation, V20, the K value increases by
11.6 K, indicating that this parameter has even more importance in
the ventilation of substation TS02 than in TS01. In simulation V26
the head loss coefcient of both inlet and outlet expanded metal
grilles has been doubled and the increase is only by 1.5 K, indicating
that, like for substation TS01, this is not a signicant parameter.
The variation of the surface area of the protruding ventilation
vents of substation TS02 can be obtained in two ways: by changing
the perimeter of the aperture over which the protruding vent lies or
by changing the number of louvers. These possible variations are
graphically shown in the pictures in Fig. 11.
In simulations V27, V28 and V29, the perimeter of the inlet,
outlet and both vents, respectively, were reduced, resulting in a 15%
decrease of the surface area. Fromthe results shown in Fig. 10, it can
be said that for substation TS02, unlike substation TS01, the varia-
tion in the K value is the same when the inlet or the outlet vents are
reduced because their number and surface areas are equal. Simu-
lations V30, V31 and V32 correspond to a 45% increase in the sur-
face area of the inlet, outlet and both ventilation vents, respectively,
which was obtained by increasing the perimeter of the aperture.
Again, the reduction in the K value is the same when the inlet or the
outlet vents are varied. Moreover, as in substation TS01, when both
inlet and outlet surface areas of the ventilation vents are modied,
the increase or the decrease in the K of the substation is the sum of
the increases or the decreases obtained with each modication
separately.
If the results of simulations 27 to 32 are compared for the two
substations, it can be concluded that the variation of the surface
area of the ventilation windows or vents has more inuence on the
enclosure class of substation TS02 than on substation TS01.
Inthe last set of simulations of substationTS02, the variationof the
ventilation surface area was achieved by changing the number of
louvers in the protruding vents. In simulations V33, V34 and V35, the
number of louvers is reduced from 8 to 4, in the inlet, outlet and in
bothprotrudingvents, respectively. Inthesesimulations there is alsoa
modicationinthe headloss coefcient of the expandedmetal grilles,
but as pointedout previously its inuence is negligible. If theresults of
these tests are compared with the results of tests V27, V28 and V29, it
canbestatedthat theperimeter of the vents has moreinuenceonthe
enclosure class of the substation than does the number of louvers: in
tests V33eV35 the reduction of area is 50%, and in tests V27eV29 it is
only 15%; however, in tests V33eV35 the increase in the K value with
respect to tests V27eV29 is only by about 1 K.
In tests V36, V37 and V38, the number of louvers is increased
from 8 to 12, in the inlet, outlet and in both protruding vents,
respectively. In this case the decrease in the enclosure class of the
substation is very small, indicating that the original number of
louvers of the protruding vents can be considered an optimum
value: the small reduction of the K value does not justify the in-
crease in the cost when moving from 8 to 12 louvers.
By comparing the results of simulations V30eV32 with V36e
V38, it can be concluded again that it is more effective to increase
the ventilation surface area by increasing the perimeter of the vents
than by increasing their number of louvers: with almost the same
surface area variation, 45e50%, in simulation V32 the reduction of
K is by 12.7 K and in simulation V38 it is only 2.6 K.
8. Conclusions
An algebraic zonal thermal model of ventilation in underground
transformer substations has been presented. The analysis of the air
ow pattern and temperature distributions obtained by a more
detailed CFD model [9] has served to develop and adjust the
simplied model proposed here.
The results of the model in the simulation of eight temperature
rise tests under different conditions are compared with those ob-
tained from the CFD model and from the experimental tests in
order to validate it. An acceptable degree of similarity has been
obtained for most of the simulations, with differences in the results
of the zonal model being less than 10% in the air mass ow rate, 1 K
in the enclosure class, and around 2 K in the outow air and
maximum inside air temperature rises.
The model has been implemented in user-oriented simulation
software that can be used by the manufacturers of substations as a
design and optimisation tool to gauge the inuence of the variation
of the geometric parameters in the ventilation and the thermal
performance of the substation.
From the parametric study carried out in order to show the
potential of the proposed zonal thermal model, the following
valuable conclusions can be inferred:
v Increasing the dimensions of the enclosure does not produce a
signicant reduction in the K value.
v It has been determined that the pass area between the LVeMV
zone and the transformer zone is one of the most critical pa-
rameters affecting ventilation. In the modelled substations this
area cannot be increased due to geometrical restrictions, but
this result can be used to design future substations with the
maximum possible pass area.
v The reduction of the head loss coefcient of the expanded metal
grilles (the grilles with the higher coefcient) does not reduce
the K value of the substation by a signicant amount.
v For substation TS01, the most effective way of reducing the
enclosure class is by increasing the surface area of the inlet and
outlet ventilation grilles.
v For substation TS02, in order to reduce the K values, it is more
effective to increase the surface area of the inlet and outlet
ventilation vents by increasing the perimeter of the aperture
than by increasing the height and the number of louvers.
The simplied zonal model offers quick thermal results with a
higher computational efciency than the previous CFD model,
making the design stage of a new transformer substation shorter.
Fig. 11. Aperture of the protruding vents (left), original protruding vent with 8 louvers (centre) and modied protruding vent with 4 louvers (right).
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 227
Using this simplied model, a complete optimisation study, taking
into account economic issues, could be proposed.
Acknowledgements
This research has been carried out as part of the research project
called CRISALIDA, with the support of Grupo Ormazabal-Velatia.
CRISALIDA was funded by Ministerio de Industria, Turismo y
Comercio through the CENIT (Consorcios Estratgicos Nacionales
en Investigacin Tcnica) program. The nancial support of Ctedra
Fundacin Antonio Aranzbal-Universidad de Navarra is also
gratefully acknowledged.
References
[1] IEC 60076-2:1997, Power Transformers. Part 2: Temperature Rise, second ed.,
IEC Standard, 1997.
[2] IEC 62271-202:2006, High-voltage Switchgear and Controlgear. Part 202:
High Voltage/Low Voltage Prefabricated Substations, IEC Standard, 2006.
[3] W.M.M. Menheere, Transformer stations and natural ventilation, in: CIRED
13th International Conference on Electricity Distribution, Brussels, 1995. Pa-
per no. 1e23.
[4] Z. Radakovic, S. Maksimovic, Non-stationary thermal model of indoor trans-
former stations, Electr. Eng. 84 (2002) 109e117.
[5] Z. Radakovic, S. Maksimovic, Dynamical thermal model of oil transformer
placed indoor, in: CIRED 20th International Conference on Electricity Distri-
bution, Prague, 2009. Paper no. 0304.
[6] I. Iskender, A. Mamizadeh, An improved nonlinear thermal model for MV/LV
prefabricated oil-immersed power transformer substations, Electr. Eng. 93
(2011) 9e22.
[7] N. Loucaides, Y. Ioannides, V. Efthymiou, G.E. Georghiou, Thermal modeling of
power substations using the nite element method, in: 7th Mediterranean
Conference and Exhibition on Power Generation, Transmission, Distribution
and Energy Conversion, 7e10 November 2010. Agia Napa, Cyprus, Paper no.
MED10/188.
[8] J.C. Ramos, A. Rivas, J.M. Morcillo, Numerical thermal modelling of the natural
ventilation of a half-buried transformer substation using CFD techniques, in:
R. Bennacer (Ed.), Progress in Computational Heat and Mass Transfer, vol. II,
Lavoisier, Paris, 2005, pp. 929e934.
[9] J.C. Ramos, M. Beiza, J. Gastelurrutia, A. Rivas, R. Antn, G.S. Larraona, I. de
Miguel, Numerical modelling of the natural ventilation of underground
transformer substations, Appl. Therm. Eng. 51 (2013) 852e863.
[10] M. Musy, E. Wurtz, F. Winkelman, F. Allard, Generation of a zonal model to
simulate natural convection in a room with a radiative/convective heater,
Build. Environ. 36 (2001) 589e596.
[11] J. Stewart, Z. Ren, COwZ e a subzonal indoor airow, temperature and
contaminant dispersion model, Build. Environ. 41 (2006) 1631e1648.
[12] J. Gastelurrutia, J.C. Ramos, G.S. Larraona, A. Rivas, J. Izagirre, L. del Rio, Nu-
merical modelling of natural convection of oil inside distribution trans-
formers, Appl. Therm. Eng. 31 (2011) 493e505.
[13] J. Gastelurrutia, J.C. Ramos, A. Rivas, G.S. Larraona, J. Izagirre, L. del Rio, Zonal
thermal model of distribution transformer cooling, Appl. Therm. Eng. 31
(2011) 4024e4035.
[14] F.P. Incropera, D.P. De Witt, T.L. Bergman, A.S. Lavine, Fundamentals of Heat
and Mass Transfer, sixth ed., John Wiley & Sons, New York, 2006.
[15] G.N. Walton, A new algorithm for radiant interchange in room loads calcu-
lations, ASHRAE Trans. 86 (2) (1980) 190e208.
[16] B.R. Hutchinson, G.D. Raithby, A multigrid method based on the additive
correction strategy, Numer. Heat Transfer 9 (1986) 511e537.
M. Beiza et al. / Applied Thermal Engineering 62 (2014) 215e228 228