Energy 111 (2016) 211e225

Contents lists available at ScienceDirect

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

Experimental and numerical investigation of a backside convective

cooling mechanism on photovoltaic panels
S. Ni
c a, *, F. Grubi
si
zeti

c- Cabo a
, I. Marini
c-Kragi
c b, A.M. Papadopoulos c
a
University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, Laboratory for Thermodynamics and Energy
Efficiency, R. Boskovica 32, 21000 Split, Croatia
b
University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, Department of Numerical Modeling and Computer
Application, R. Boskovica 32, 21000 Split, Croatia
c
Aristotle University of Thessaloniki, Department of Mechanical Engineering, Process Equipment Design Laboratory, GR-54124 Thessaloniki, Greece

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, two generic photovoltaic (PV) panels (poly-Si and mono-Si) were experimentally tested in
Received 29 January 2016 typical Mediterranean climatic conditions. The focus of the applied experimental approach was to
Received in revised form examine the effect of backside convective thermal profile and its impact on temperature distribution, i.e.
24 May 2016
on panel electrical efficiency. Therefore, a series of measurements was made in 2015, from April to July, as
Accepted 26 May 2016
Available online 5 June 2016
well as CFD modeling in order to obtain a detailed analysis of the possible working regimes. According to
the obtained experimental and CFD results, the present design of typical PV panels have an unfavorable
impact on PV panel electrical efficiency. Namely, typical contemporary panel designs lead to two typical
Keywords:
Photovoltaics
backside convective air temperature profiles which have a direct impact on the effectiveness of natural
Renewable energy cooling. As shown in the obtained measurements, the specific convective profiles at the backside of PV
CFD analysis panels have a significant influence on the degradation rate of panel electrical efficiency in the estimated
Experimental analysis amount of 2.5% up to 4.5%. The results of the research discussed in this paper signal the need to provide a
Energy efficiency possible redesign of the backside surface in conventional PV panels, in order to increase their average
Thermodynamics efficiency (more efficient backside thermal dissipation).
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction
Sufficient production of electricity from photovoltaic systems
strongly depends on the average achieved electrical efficiency of
photovoltaic panels (PV), i.e. PV systems, apart from the specific
climate circumstances given in which a PV system operates. The
most widely used contemporary photovoltaic technologies are
siliceous based ones, mono-or polycrystalline, [1,2]. Despite the
significant progress that has been made since the early 1990s,
raising electrical efficiency in PV systems and reducing their pro-
duction cost are still major research goals, both in basic and applied
ones, especially when considering the relatively high overall in-
vestment in PV systems [3].
One possible way of achieving increase in PV efficiency is by
applying a financially feasible cooling technique that can lead to the
decrease of panel temperature and therefore increase its electrical
* Corresponding author.
E-mail address: snizetic@fesb.hr (S. Ni

zeti
c).
efficiency. There are several research studies related to the inves-
tigation of different cooling techniques and their effect on panel
performance, [4e10]. As these studies showed, there is a potential
of increasing PV efficiency in that manner, but the achieved in-
crease is generally modest which questions the feasibility of certain
cooling techniques. In most of these studies, emphasis was placed
on technical issues and the economic aspects were linked to them.
A comparison of different cooling techniques for PV panels in
relation to the obtained relative increase in specific panel power
output can be found in Ref. [5].
The development of efficient hybrid PVT systems is also
important in order to enable a more efficient utilization of existing
PV technologies and also to penetrate into the market more deeply.
The main advantage of PVT systems is the possibility of simulta-
neous heat and electricity production, whereby the total achieved
efficiency of the considered PVT systems can be relatively high. In
the specific topic of PVT systems, research efforts are considerable
and different PVT concepts have been investigated in recent years,
[11e16].
Another option that can lead to the increase of overall electricity

http://dx.doi.org/10.1016/j.energy.2016.05.103
0360-5442/© 2016 Elsevier Ltd. All rights reserved.
212 S. Ni
zetic et al. / Energy 111 (2016) 211e225
production in PV systems is to provide maximization of power
output through a suitable MPP tracking strategy [17e22], which
needs to be economically viable. Technically speaking, this does not
lead to an increase in panel electrical efficiency as such; the effi-
ciency increase is due to prolonged time exposure at more effective
solar radiation angles. Moreover, MPP tracking technologies have
significantly improved, and their costs have decreased drastically in
the 2010s and have become a very attractive investment option.
Novel photovoltaic technologies that are more efficient than
currently available ones are still under research. The main problem
generally lies in their economic aspect. It is therefore clear that
emphasis needs to be placed on novel, more efficient photovoltaic
technologies that are cost-effective. However, the improvement of
existing PV technologies should not be neglected, as it would be
much easier to push modified and improved PV systems to the
market than a new technology that may lead to reservations on
behalf of investors.
The main objective of this paper is to examine the effect of the
convective thermal profile that occurs on the backside of photo-
voltaic panels and of its overall impact on panel backside temper-
ature distribution, i.e. on electrical efficiency. Besides an
experimental approach, CFD modeling was also used and the re-
sults obtained were compared to the experimental readings, a task
accompanied by a sensitivity analysis in relation to the PV panel
temperature distribution. The main research outcome of the herein
presented study can lead to the consideration of PV panel backside
surface redesign in order to achieve an increase in panel electrical
efficiency.
2. Theoretical background basis
Photovoltaic cells gain energy directly from solar irradiance
where just one part of incident irradiation is converted into useful
electrical energy by means of the photoelectric effect. The rest of
the incident solar energy is mostly accumulated on the PV panel by
absorption as heat gain; a smaller portion is reradiated to the sur-
roundings. Since the ambient is almost always at a lower temper-
ature than the PV cell, heat tends to dissipate into the surroundings.
The intensity of heat dissipation varies due to many parameters
such as air flow (direction and magnitude), average air temperature
and surrounding relative humidity, bracket shape, reflectivity of
surrounding surfaces, etc., i.e. it is certainly a complex issue. For the
purpose of this analysis, a simplified theoretical overview is pro-
vided, which briefly explains energy fluxes through the PV cell and
is useful to understand the root of possible impacts on panel per-
formance. Basic heat fluxes are presented in Fig. 1 for the case of
typical PV panels.
In steady-state conditions, the incoming energy is equal to the
energy dissipation from the considered system presented in Fig. 1,
E_ in ¼ E_ out . Since all energy flows originate from incoming solar
irradiance, direct insolation is significantly greater than diffuse
radiation at peak heat gains. The heat income can be defined as
follows:

  
a$Gs $Ap ¼ Gs $Ap $hcell þ Ap $afront $DTeq: þ Ap $aback $DTeq: back þ
h
i h

front i
þ ε$s$Ap $ Tpanel
4 4
 Tamb: þ ε$s$Ap $ Tpanel 4 4
 Tamb
front
Q_ in ¼ E_ in ¼ a$Gs $Ap ; (1)
where a presents the overall solar absorption coefficient for the
whole PV cell.
A standard PV panel usually consists of glass, silica, ethylene
vinyl acetate (EVA) and polyvinyl fluoride (PVF) layers (Tedlar layer
in Fig. 1). As every material has a different absorptivity coefficient, it
is difficult to accurately predict the amount of absorbed heat.
Moreover, it is also a complex issue to measure the spectrum of
irradiance, in order to show which wavelengths are emitted from
the sun, and what fraction of each wavelength is transmitted
(absorbed) into different layers of the PV panel. Although the pre-
viously mentioned information can be computed, it strongly de-
pends on atmospheric conditions such as humidity, overcast and
the amount of dust in the air. Time of day is also in direct correlation
with the shape of irradiance spectrum.
Overall PV cell heat output cannot be strictly defined, as it de-
pends on cell efficiency as well as on surrounding circumstances in
which the PV system operates and finally from the cell material.
Further, overall PV panel energy output can be written as follows,
E_ out ¼ Q_ heat þ E_ current (2)
where Q_ heat is heat rejected from the PV panel in steady-state
conditions to the surroundings and E_ current is the produced elec-
trical output. Overall PV panel rejected heat is due to convection
and thermal radiation, respectively,
Q_ heat ¼ Q_ convection þ Q_ radiation : (3)
Heat dissipation to the surroundings as a result of convection
depends in general on a series of factors such as tilt angle, material
type, surrounding air speed and temperature, average panel surface
temperatures, time of day, etc. Besides temperature, thermal radi-
ation depends upon the emissivity and reflectivity of the material.
Furthermore, emissivity defines how much of the heat will be
dissipated to the surroundings, while reflection defines how much
of the surrounding heat will transfer to the PV cell from the back
side and lateral frame of the PV panel. As PV panel temperature is
almost always higher than the temperature of surrounding air, it
can be concluded that dissipation will always be greater, in com-
parison with diffuse insolation absorption, mainly the reflected
insolation from surrounding surfaces back to the PV panel. Never-
theless, radiation intensity is strongly correlated to the ambient
temperature and reflectivity of surrounding surfaces.
Therefore, a basic heat equation (energy balance equation) can
be written for the considered PV panel in steady state conditions,
Q_ heat ¼ E_ current þ Q_ convection þ Q_ convection
frontside backside
þ Q_ radiation _
frontside þ Q radiation backside (4)
(5)
back
 S. Ni
zetic et al. / Energy 111 (2016) 211e225
Fig. 1. Elementary energy flows through the PV panel.
In the previously elaborated theoretical overview, it is notice-
able that different influential factors can affect PV performance.
Hence, in realistic operating circumstances, a major issue is the one
related to heat transmission mechanisms that occur over the PV
panel surface, and that strongly depend on the surrounding con-
ditions. Heat convection and thermal radiation are approximately
in the same order of magnitude; however, we focused on convec-
tive thermal influence in this study. Namely, if we take the average
measured parameters into account as well as other well-known
physical characteristics for the photovoltaic panel, i.e. average
floor temperature (30  C), average PVF layer temperature (48  C),
average glass temperature (40  C), average surrounding air tem-
perature (28  C), clear sky temperature average, view factors,
insolation of around 860 W/m2, PVF emissivity (0.86), glass emis-
sivity (0.9), wind velocity of around 1.0 m/s, we can estimate the
amount of convection and thermal radiation heat flux, i.e. overall
heat flux. Therefore, using the well-known standard thermody-
namic relations (eq. (5)) and the previously specified average
values, the overall calculated convection heat flux is at around
350 W/m2 and the thermal radiation flux is at around 260 W/m2
(overall at around 610 W/m2). Thus in the analyzed example,
around 57% of heat transfer is due to heat convection and the rest
due to thermal radiation (we did not take the frame into account in
this approximate calculation). According to the provided estima-
tion, it is clear that thermal radiation cannot be neglected. Hence, in
our CFD modeling approach, we later took the influence of thermal
radiation into account but in an indirect way and it will further be
discussed in section 4.1.
Since PV electrical efficiency significantly depends on its oper-
ating temperature, it is important to describe the cooling mecha-
nisms that occur over the PV panel surface. According to Teo et al.
[8], in most of the developed cases, numerical models predict
higher values of the temperature inside the PV cell in comparison to
the real values. This issue is understandable, considering the fact
that glass is mostly transparent in a wide range of the solar radia-
tion's spectrum, which means that PV cells are directly heated from
the Sun. Furthermore, it is also important to emphasize the prob-
lem of dirt and grease accumulated on panel surfaces which not
only block incoming irradiance, but also unnecessarily heat the
front surface of PV panels due to the opaqueness of the impurities.
Putting the physical process of the photoelectric effect aside, a
PV cell can be viewed as a heat source in a pure thermodynamic
aspect, where main heat dissipation occurs at the front and back
side, at glass and PVF back plate surfaces respectively. Lateral sides,
often made from aluminum, have significantly higher heat dissi-
pation capacity. Most of the cells end up insulated from the
aluminum frame and results in dominant heat flux at the front and
back side of the PV panel.
213
Because of its electrical properties, PVF has rather low thermal
conductivity properties and acts as an insulator: the heat conduc-
tivity coefficients PVF and glass are 0.2 W/mK and 1.5 W/mK
respectively. Since the thickness of PVF and glass layers are around
0.3 mm and 3.0 mm respectively, it can be deduced that the heat
dissipation from the front and back side of the PV panel is of the
same order in magnitude (if we assume that the convective heat
transfer coefficient to surrounding air is roughly the same on the
front and back side of the PV panel).
PV cell temperature is mostly constant through single cells,
because of its high conductivity that is around 150 W/mK and can
exceed 60  C, this depends on climate conditions. It is known that
cell efficiency drops with its temperature at a rate of about
0.45e0.5% per degree of Celsius [23], depending on the type of PV
technology. The cells are not directly connected to the aluminium
frame, so heat dissipation tends to slow down from the edge to the
centre of the PV panel. Hence, in this line of approach, there is a
temperature increase in centrally fixed cells on the PV panel, which
degrades its electrical characteristics, i.e. its electrical efficiency
(infra-red solar irradiation, which passes through the glass and PV
cell, gets absorbed directly into the PVF material, however, this
amount of heat is nearly impossible to measure and define, so its
influence in the herein presented analysis is neglected as already
seen in Ref. [8]).
Interface resistance between solid/solid and solid/fluid is
important for thermal phenomena and interface resistance could
be a potential source of error which is currently being studied, [24].
Finally, it is important to obtain temperature distribution over
the PV panel as an important ground for deducing the conclusions
related to the possible redesign of conventional PV panels, in order
to dissipate as much heat as possible. This paper focuses on
convective heat transfer issues that occur on the back side of PV
panels.
3. Experimental approach
In order to obtain temperature distribution over PV panel sur-
faces and examine heat transfer issues, an experimental approach
was performed. Experiments were done from May to July 2015 in
Split, Croatia, which has a typical Mediterranean climate [25]. The
experiments were conducted every 10e15 days in order to have a
valid sample for data analysis.
3.1. Description of specific experimental setup
A schematic depiction of the experimental setup is presented in
Fig. 2, a photo of the assembled experimental setup during mea-
surement preparation is shown in Fig. 3. Two commonly used types
of PV panels were used, poly-crystalline and mono-crystalline Si-
based PV panels, each with a nominal power output of 50 W; their
main technical data are presented in Table 1. The mono-crystalline
PV panel dimensions were 550  650 mm; the poly-crystalline
dimensions were 650  600 mm and were fixed at an angle of
20 respectively. The angle was chosen with respect to the
geographical location where the measurements were performed.
Both examined PV panels were measured under the same condi-
tions to enable a comparison of results. A pyranometer (HAENNI
model Solar 130) was used to detect the amount of incoming solar
irradiance and both panels were connected to a power load (45 W
halogen lamp) in order to obtain realistic working conditions.
The air velocity and ambient air temperature were monitored
throughout the period of measurements. To obtain exact cell
temperature, the PVF layer was purposely drilled and a thermo-
couple sensor (TC) was placed directly on the siliceous cell. A total
of four thermocouples were used to measure the temperature of
214 S. Ni
zetic et al. / Energy 111 (2016) 211e225

Fig. 2. Simplified schematic of the experimental measurement ring.

also used their TC-08 temperature data logger with appropriate

Fig. 3. Photo of the experimental setup during measurement preparation.
four cell series in each PV panel. Furthermore, two thermocouples
were used to directly measure the PVF surface temperature. In or-
der to measure real layer temperature rather than mean temper-
ature between the back side of the PV panel and surrounding air,
the thermocouples on the back side were covered with a tape that
has similar properties as the PVF. Two thermocouples were used to
measure the temperature of the glass above and two thermocou-
ples were used to measure the temperature on the back side of the
PV panel. The thermocouples used were from PICO (K-type); we
data acquisition software (8 channel thermocouple data logger,
measurement range being from 270  C up to 1820  C, with an
accurate fast sampling rate and high resolution). Additional sensors
were used to measure temperature and relative humidity of the air
simultaneously on the backside surface of the PV panel, as shown in
Fig. 4; they were CAP sensors, named as Sensirion EK-H4 mea-
surement kit (plug-and-play system connected to the PC via USB for
simultaneously measurement of temperature and air humidity for
up to four measurement channels). These sensors are especially
suitable for air temperature and humidity measurement when it
comes to temperatures for up to 100  C. For measurements of
average air velocities, a hot-wire thermo-anemometer was used
(Kimo VT-100, thermos-anemometer for the measurement of air
velocity, air flow and surrounding air temperature, equipped with a
standard 306 mm long  8 mm diameter probe and 2 m in length).
Finally, the main goal was to detect the temperature in the
boundary layer of the back surface, i.e. to detect the type of specific
convective thermal profile in the air, for typical PV panel operating
circumstances in the Mediterranean. The same approach was
applied for both mono-crystalline and poly-crystalline PV panels.
The positioning of the previously described temperature sensors is
presented in Fig. 4 for the front and backside of the PV panel.
Several parallel measurements were obtained between April
29th and June 16th, 2015. This was done between 11.00 and 14.00,

Table 1
General characteristic for the considered PV panels (mono and poly).

Basic technical data for PVa panels

Model SL-50AA36 SL-50P

Number of cells in the module 36 36
Maximal power output 50 W ± 3% 50 W ± 3%
Maximal power voltage/Open circuit voltage 17.2 V/22.9 V 18.0 V/22.5 V
Maximal power circuit current/Short circuit current 2.91 A/3.18 A 2.78 A/3.02 A
Panel area 0.357 m2 0.39 m2
a
PV panel characteristics at irradiance of 1000 W/m2 and 25  C.
 S. Ni
zetic et al. / Energy 111 (2016) 211e225 215

Fig. 4. Specific positioning of temperature sensors: (a) backside of the PV panel, (b) front side of the PV panel.

InsolaƟon: 906 W/m

Air velocity: 1.0 m/s
55 Temperature of the surrounding air: 21°C

50

TC Cell 1

TC Cell 2
45
TC Cell 3
Temperature (°C)

TC Cell 4

TC Tedlar 1
40
TC Tedlar 3

TC Glass 1

TC Glass 3
35
TC Frame

30

25
0 20 40 60 80 100 120 140 160 180 200

Time (s)

Fig. 5. TC temperature distribution for the poly-crystalline panel (April 29th).

as the aim was to monitor the period of highest solar irradiation
levels where backside convective thermal exchange is most
intensive. The results varied, mostly depending on wind velocity
and surrounding air temperature, time of day and irradiation in-
tensity levels. Typical temperature distribution profiles for the
considered PV panels are presented in Figs. 5 to 8. The aluminum
frame temperature was measured with additional installed ther-
mocouples and the data shown in the upcoming figures are average
values for the frame's temperature (surrounding air temperatures
are also averaged).
In the next section of the paper, we discuss a detected backside
convective thermal profile of the air that occurs on the backside
surface of the PV panel in certain circumstances (in relation to that,
we also address the above presented figures).
3.2. Characteristic convective heat transfer at the backside surface
of the PV panel
The backside surface of PV panels is influential regarding
desired heat rejection, so we wanted to investigate these heat
216 S. Ni
zetic et al. / Energy 111 (2016) 211e225

42
InsolaƟon: 906 W/m
Air velocity: 1.0 m/s
Temperature of the surrounding air: 21°C
40

38
Temperature (°C)

36

34

32

30
0 20 40 60 80 100 120 140 160 180
Time (s)

Fig. 6. CAP temperature distribution for the poly-crystalline panel (April 29th).

InsolaƟon: 869 W/m

Air velocity: 1.0 m/s
Temperature of the surrounding air: 28.3°C

60

TC Cell 1
55

TC Cell 2

TC Cell 3

TC Cell 4

50 TC Tedlar 1
Temperature (°C)

TC Tedlar 3

TC Glass 1

TC Glass 3
45
TC Frame

40

35
0 20 40 60 80 100 120 140 160 180

Time (s)

Fig. 7. TC temperature distribution for the mono-crystalline panel (July 9th).

transfer circumstances both experimentally and by using CFD occurrence of the specific thermal profile as a result of that.
simulations in order to cover a wider range of possible working Furthermore, backside convective thermal profile presents an
regimes, provide a sensitivity analysis and confirm the physical important feature as it acts as a cooling mechanism for PV panels
effects that were sighted. and generally defines the amount of rejected heat from the PV
panel. According to the provided experimental measurements, we
3.2.1. Backside convective air temperature profile detected two typical convective temperature profiles at the back-
The backside surface of PV panels is proven to be warmer than side surface of the PV panel; they are presented in Fig. 9. Temper-
the surroundings in most cases and convection leads to the ature air profiles are presented in relation to the distance measured
 S. Ni
zetic et al. / Energy 111 (2016) 211e225
45
44
43
42
41
Temperature (°C)
40
39
38
37
36
35
0 20 40 60
Fig. 8. CAP temperature distribution for the mono-crystalline panel (July 9th).
Fig. 9. Specific types of characteristic convective air and PV cell temperature profiles
on the backside surface of the PV panel.
starting from the front lower part of the PV panel; the position
mentioned corresponds to the specific measurement points of the
thermocouple sensors, as shown in Fig. 4a and b. The first specific
case is convective thermal profile, with red circles in Fig. 9 (the
curves in Fig. 9 are approximately fitted in order to show the curve
trend), Case I, where air thermal profile temperature gradually
decreases from the bottom section of the PV panel towards the top
section of the PV panel (approximately section A, Fig. 4). This means
that the bottom sections of the PV panel, for the case considered,
are generally exposed to the highest temperatures, which means
that in the mentioned sections of the PV panel, the cell's electrical
efficiency degradation is at its highest and certainly higher
compared to other sections of the PV panel, in the middle and top
section, i.e. respectively section B and section C in Fig. 4. According
to the experimental results, the mentioned profile of backside
convective thermal profile mainly occurred in spring where sur-
rounding air temperature was lower in comparison to summer.
However, solar irradiation levels and surrounding air velocities
217
InsolaƟon: 869 W/m
Air velocity: 1.0 m/s
Temperature of the surrounding air: 28.3°C
CAP
Sensor 1
CAP
Sensor 2
CAP
Sensor 3
CAP
Sensor 4
80 100 120 140 160 180
Time (s)
were approximately the same, so the temperature of surrounding
air is a crucial factor. In summer period, we noticed a second case
(CASE II, Fig. 9) which is especially interesting because it is notably
different from the first one (CASE I, Fig. 9), i.e. different in relation to
the expected convective temperature profile at the backside of the
PV panel. Namely, air temperature is again highest at the bottom
sections of the PV panel in the second noticed case. However, air
temperature (section B, Fig. 4) is lower in the middle section of the
PV panel in comparison with the upper section of the PV panel
(which was opposite according to expectations). Namely, it is ex-
pected that air temperature gradually decreases from the bottom
section of the PV panel towards the top backside section of the PV
panel. The specific temperature profile is presented in Fig. 9 in
green triangles. In the considered case, the air temperature at the
backside section of the PV panel was around 43  C (section A),
around 36.7  C in the middle section (section B) and about 40  C at
the top backside section (section C). The noticed effect has direct
implication on the increased degradation of PV panel electrical
efficiency on its entire area, as all sections are exposed to higher
temperatures in general. According to the experimental experience,
the effect occurred during the highest solar insolation levels and
relatively weak wind velocity which was around 1.0 m/s and
certainly not higher than 1.5 m/s. In the case of higher surrounding
air velocity magnitudes, this effect remained but with a lower
overall PV panel temperature due to an increased heat rejection
rate. Additionally, we noticed that in the second case of specific
backside convective thermal profile the difference between cell
temperatures in section A and section C ranged between 6.0  C and
9.0  C, whilst in the first mentioned case, this difference ranged
between 2.0  C and 4.0  C. Hence, as a result of general higher cell
temperatures, it is obvious that we can certainly expect an increase
in average cell temperature ranging approximately from 3  C to 6  C
in the second case considered.
However, the previously mentioned cell temperature difference
strongly depends on stochastic surrounding conditions, especially
when it comes to surrounding wind magnitude and direction. As it
was noticed, the second discovered profile of convective backside
thermal profile was particularly evident in periods with high solar
218 S. Ni
zetic et al. / Energy 111 (2016) 211e225
insolation levels and relatively weak prevailing air velocity.
Therefore, due to a relatively weak convection heat rejection rate
for the backside sections of the PV panel and also due to relatively
high solar insolation levels, it can be presumed that a ‘’cloud’’, i.e. a
flow separation of warm air accumulates at the lower and upper
edge of the backside sections of the PV panel creating a kind of hot
air bubble. The hot air bubble insulates the backplane from ambient
air, likely leading to the degradation of PV efficiency; in comparison
to the first case (previous statement is later confirmed in the ob-
tained CFD analysis). The second effects (CASE 2) occur in the worst
possible case, as higher solar insolation levels are available, but are
unfortunately followed by highest panel electrical efficiency
degradation, which has a direct effect on the amount of PV system
produced electricity. In both presented cases, the insolation was
around 868 W/m2, the average surrounding air temperature was
around 29  C and the average aluminum frame temperature at
around 37  C. Hence, from the previous elaboration, the importance
of specific backside convective thermal profile becomes obvious as
it has direct implications on panel electrical efficiency (because of
increased average PV panel temperature). It is clear that different
surrounding conditions can affect the backside temperature profile.
3.2.2. Temperature profiles on backside surface
As it can be noticed in Figs. 5 and 7, the PVF layer temperature
tends to be higher than the cell temperature in certain cases, which
was not expected and should not be the case in theory according to
the material properties and according to well-known thermody-
namic laws. To remove any doubt, the thermocouples were checked
after measurements in order to ascertain their efficiency. It essen-
tially seems that even though a PV cell presents a heat source, the
temperature at the back side of some cells is substantially higher, by
up to 3.0  C, than the temperature of the PV cell itself in the same
time frame. One possible explanation could lie in additional
reflective insolation onto the back side from the ground, but was
also eliminated by reducing the reflectivity of the ground with the
use of dark cloths. Nevertheless, the effect remained unchanged.
The only viable explanation lies in the higher amount of infrared
irradiance (with wavelengths being higher than 1.1 mm in summer),
which partially penetrates the front glass layer, silicon PV layer and
PVF layer and directly heats the thermocouple probe, which is
completely opaque for IR irradiance (a similar effect was mentioned
in Ref. [8] and was herein experimentally proven). The main indi-
cator is the fact that the effect is more visible in summer, when the
amount of infrared irradiance should be much higher. The easiest
way to prove this should be to make measurements in winter, when
there should be less infrared irradiance and that could be a good
basis for reaching a conclusion beyond doubt. Presumably, this
effect was not perceived in other research studies as other mea-
surements were done with uninsulated thermocouples. On the
other hand, the thermocouples used in this study were insulated
with materials similar to PVF, in order to evade measuring mean
temperature between the ambient air and PVF layer. Transmission
coefficients for wavelengths above 1.1 mm are around 0.52, 0.92 for
PV panel materials and 0.95 for siliceous, PVF and glass layers
respectively, hence the explanation which takes into account a high
amount of infra-red irradiance could prove viable.
3.2.3. Specific temperature profiles: correlation of PV cell
temperature profile with air
temperature profile at backside of PV panel.
Temperature distribution patterns for PV cells were discussed
earlier and considering most heat gets away through the aluminum
frame, convection heat transfer is roughly the same across the
whole PV panel and the highest heat values should be somewhere
in the middle of the panel. However, the experimental results
showed a certain difference: when comparing temperature read-
ings, as shown in Figs. 5, 6 and 9, it can be noticed that no matter
cell number 4 being low in temperature, the surrounding air tem-
perature behind the cell is fairly high. A similar effect is noticed
when comparing sensor 1 and cell 1 temperature, Figs. 7 and 8.
Further on, the same effect is also visible when comparing readings
from temperature sensor 3 and cell 3, although it needs to be said
that sensor 3 is much closer to the edge of the PV panel when
compared to cell 3. Different conclusions can be drawn when we
compare readings from the previously mentioned sensors with
readings obtained by the rest of the sensors. However, it seems that
the air temperature around the back edges of the PV panel does not
correspond to the thermal profile of the whole PV module (Fig. 9).
Also, by looking at Fig. 9, it can be seen that the lower part of the
module (region A) operates on a higher temperature than the rest
of the cell, especially higher than the upper part (region C). Pre-
sumably, the cause of high temperature in region A is due to the low
convection heat exchange of the considered section, which can be
explained with flow separation i.e. the forming of an air bubble
underneath the panel. If we take into account that the temperature
in region A is 6  C higher than the temperature in region C, and that
efficiency degradation changes by 0.5%/ C, we can conclude that
region A has a lower efficiency by 3.0%. Since region A covers one
third of the panel, we can conclude that the electrical efficiency of
the PV panel dropped by around 1.0% only as a result of the bracket
design.
3.3. Measurement error analysis
In this section we briefly elaborate the influence of expected
measurement error. The expected measurement error for the
applied measurement equipment used in the experiment (i.e.
guaranteed measurement error) was specified in Table 2, according
to available technical data sheets of the equipment producers. Ac-
cording to the specified values in Table 2, the highest measurement
errors could be expected for the case of air velocity measurement;
they could be expected to be around ±3%. However, regarding
realistic ranges of air velocities during measurements, the deviation
for the measured results is not significant.
Other used sensors in the herein presented experiment, i.e.
temperature and solar insolation sensors, have a relatively modest
guaranteed measurement error, so their influence is also rather
insignificant. Finally, the overall expected influence of guaranteed
measurement error in temperature measurements, the most
important parameter for the purpose of this analysis, is around
±0.3%, which means that the general impact of guaranteed mea-
surement error is minor and does not have a significant impact on
the purpose of this study.
In order to visualize air and thermal flow as well as obtaining a
sensitivity analysis, a CFD analysis for mono-crystalline PV panels
was carried out.
4. CFD analysis and comparison with experimental results
A PV panel might appear to be a simple device; however, it in-
volves relatively complex heat transfer processes where panel
performance significantly depends from variable surrounding
conditions. To model the mentioned processes, a computational
fluid dynamics (CFD) technique was implemented. In the case of
the PV panel being exposed to open surrounding conditions, wind
has a high turbulent intensity and this prevents making it
mandatory to model the flow as a turbulent one. Since the occur-
ring heat transfer includes a good part of convection, the correct
modeling of the region close to the wall is important. In many cases,
wall functions are sufficient and are often used as an approximation
 S. Ni
zetic et al. / Energy 111 (2016) 211e225 219

Table 2 predictions of the onset and the amount of flow separation under
Guaranteed measurement error for the used measuring equipment. adverse pressure gradients, because it considers the transport of
Sensor type Measurement turbulent shear stress [26]. Furthermore, the convective heat
Error
transfer showed relatively good agreement with the experimental
data [27] and was successfully used in the modeling of a stand-
Temperature (EK-H4) ±0.3%
alone PV system [28], which makes the SST model sufficiently
Temperature (TC-08) ±0.2%
Pyranometer ±0.6% realistic for the considered case of the PV panel being exposed to
Wind speed ±3.0% open wind conditions. In the case of low wind velocity, natural
convection effects are noticeable, as this is a free buoyancy driven
flow where air compressibility is taken into account for those cases
in sub-viscous and buffer regions near the wall. When a level of by the ideal-gas law.
accuracy beyond that is needed, a turbulence model is used to solve In addition to modeling fluid flow, conduction and convection,
the entire flow regime. In cases when flow is not fully turbulent, various additional effects could be considered. For example, a
such as in a natural convection problem, the near wall flow needs to detailed radiation environment, electrical behavior, or even
be resolved and wall functions should not be used. In the herein coupled conductiveeradiative heat transfers with included photo-
analyzed case, the PV panel is exposed to open surrounding con- electric phenomena within PV cells [29] could be implemented to
ditions, where the free stream flow is assumed to be fully turbulent, make the model more realistic. The prime objective of CFD analysis
with a high turbulence intensity of wind stream. Nevertheless, in this paper is to isolate the effect of convection heat transfer and
accurate resolving of the boundary layer is of critical importance to investigate the effect of fluid flow and convection on backboard
correctly model convection problems and possible flow temperature. While upgrading the simulation model to include
separations. multi physics effects would be interesting, it would be hard to
Two-equation turbulence models are often used for numerical isolate the effect of convection on backboard temperature. Conse-
modeling of flow since they are robust, they have shown good re- quently, only a simple CFD model is implemented so that a good
sults and only modestly expand the system of governing equations insight in fluid flow and the respective convection heat transfer
in numerical terms (continuity, momentum, and energy equations). phenomenon could be investigated precisely.
The continuity and momentum equations in Reynolds averaged Generally, the simulation of PV should not consider an isolated
form can be written in tensor notation and Cartesian coordinates system but should replicate the respective microclimate in which it
as: operates [30]. PV panel performance obviously depends on its ra-
diation environment such as shadowing effects and reflections
 
vr v ruj from neighboring buildings or nearby surfaces of different reflec-
þ ¼0 (6) tive properties. So if one wants to simulate PV system performance
vt vxj
for specific operating conditions, it is very important to have a
" ! # detailed model of the surrounding environment. Since the objec-
vrui vrui vp v vui vuj 2 vuk v   tive of this paper is to isolate the effects of fluid flow on convection
þ uj ¼ þ m þ  dij þ R
vt vxj vxi vxj vxj vxi 3 vxk vxj ij heat transfer, it is convenient to suppress the effects of radiation so
that a general conclusion on convection effects can be easily ob-
þ rfi
tained. To achieve this, radiation is not directly modeled, but only
(7) its overall effect on heat transfer is taken into account. To simulate
solar irradiance, an equivalent heat source approach was adopted: a
" ! #
vui vuj 2 vuk 2 continuous heat source was applied directly on the PV cell, the
Rij ¼ rui uj ¼ mt þ  d  rkdij (8) magnitude of which corresponds to the intensity of irradiance in
vxj vxi 3 vxk ij 3
realistic circumstances minus the produced electric power.
The specific body force fi includes external body forces such as
gravity in this case; and Rij is Reynolds stress tensor according to the 4.1. Numerical model setup
Boussinesq eddy viscosity assumption. Turbulent viscosity mt and
turbulence kinetic energy k are obtained from a selected turbulence The earlier described model is solved numerically by a general-
model. The energy conservation equation can be written as: purpose, widely used commercial CFD software, namely ANSYS
FLUENT [31e33]. The first step is the selection of the computational
" #
domain for fluid and multiple solid regions as it is illustrated in
vrE v h i v vT
þ ðrE þ pÞuj ¼ ðkc þ kt Þ þ Sh (9) Fig. 10a, illustrates inlet and outlet boundary conditions of the fluid
vt vxj vxj vxj
region and its size. The domain size was selected with respect to
E is defined by: panel length which was Lp ¼ 0.55 m. The length of the fluid domain
was set to L ¼ 9Lp, of which 3Lp domain length was from the inlet
p u2 to the panel and the domain height was H ¼ 7Lp while the width
E ¼h þ (10) was W ¼ 7Lp (see Fig. 11a and b).
r 2
A simplified scheme of the PV panel is illustrated in Fig. 10b,
In the energy conservation equation (9), viscous dissipation with further elements: glass, PV cell, backboard (PVF) and
terms are not included since they are negligible for a low Mach aluminum frame. Each element (material) that the PV panel is
number. For motionless solid regions, the second term on the left consisted of is described with its thermal properties.
hand side and kt are removed from the equation. The chosen mesh is a tetrahedron mesh with thin prism ele-
To close the governing equations, an SST model [26] was used. It ments in the near wall boundary layer and the element size near
is a combination of two other turbulence models, k-epsilon in the the PV panel is 5 mm with a 0.01 mm height of elements in the
free stream, and k-omega models near the walls. It does not use prism layer (Fig. 1). The prism layer consists of 10 elements with a
wall functions and tends to be most accurate when solving the flow growth rate of 1.15 and far away from the PV panel; the size of the
near the wall. The SST k-omega model also gives highly accurate elements is up to 150 mm with a growth rate of 1.2. The mesh type
220 S. Ni
zetic et al. / Energy 111 (2016) 211e225

Fig. 10. Specific geometry: a) fluid zone: Lp e panel length, L e domain length, H- domain height, W e domain width and b) PV panel.

Fig. 11. a) Mesh elements in fluid region, b) Prism mesh elements near PV panel surface.

Table 3 modern 4-core Intel-processor-based PC. In comparison, for tran-

Mesh elements type and number per subdomain. sient analysis when starting from the standard initialization (con-
Subdomain Number of elements Type stant velocity, pressure and temperature field throughout the
Surrounding air 1,809,844 Tetrahedrons
domain), the computational time was about 10e30 days to
PV cell 4320 Hexahedral convergence depending on the selected wind speed. For several
Backboard 86,144 Tetrahedrons conducted transient cases, the average temperature of the back-
Glass 22,663 Tetrahedrons board was monitored. In all the investigated cases, after the average
Aluminum 54,851 Tetrahedrons
temperature converged to a constant value, no transient effects
were observed, so steady state analysis was implemented for the
continuation of the study. Furthermore, the converged average
and number for the individual subdomain is given in Table 3. The
convergence criterion was set to 104 and the second order upwind
scheme was used for discretization of both momentum, energy and
turbulent equations.
The mesh for the PV panel elements is also a tetrahedral mesh
(Fig. 12) with relatively large elements, since fine mesh is not
essential for this zone.
Inlet velocity profile was set to the respective constant value (by
height). The pressure outlet was set to 101 kPa. PV panel surface
walls and ground surface were described with no-slip boundary
conditions. The remaining boundaries that enclose the computa-
tional domain were described with free-slip boundary conditions.
Initial conditions are constant fluid velocity, constant fluid and
solid temperature at 300 K.
Both transient and steady state analysis where conducted, for
various insolation, wind speed and wind angle conditions. While
transient simulation takes many days until convergence, steady
state analysis is much faster. The steady state analysis computa-
tional time was about 5 h per simulation until convergence on a Fig. 12. Mesh elements in solid region (PV panel).
 S. Ni
zetic et al. / Energy 111 (2016) 211e225 221

backboard temperature was the same in both steady state and

transient analysis.

4.2. Validation of the CFD model and comparison with experimental

readings

Various transient and steady state analyses with different

insolation, wind velocity and wind angle conditions were con-
ducted to confirm the CFD model as well as to investigate the effect
of different working regimes on panel performance. The CFD model
validation summary and comparison with experimental readings is
given in Table 4. The measured data of various experiments was
compared to numerical model results with corresponding working
conditions that are specified by average wind velocity v, wind angle
a (direction according to Fig. 19) and irradiation G.
To simulate a real-life insolation of 869 W/m2 for the purpose of
CFD simulation, we took a heat equivalent of 650 W/m2 for the heat
source that was applied on the PV panel, the heat equivalent of the
heat source was applied on the PV cell, assuming that a specific
amount of heat is absorbed into the PV panel and that this is the
amount of heat absorbed into the PV panel. We first presumed that
PV cell was mostly opaque for solar irradiance, which is not the case
for glass and PVF material (from the previous reason we had placed
heat source at the top of the PV cells). Secondly, an absorption
coefficient of 0.86 was chosen for preliminary modeling. However,
since radiation was not directly taken into account in the numerical
model, the front side of the PV panel was not analyzed in this paper.
The intensity of heat source was decreased by the amount of
calculated thermal radiation from the back side. Also, we needed to
decrease equivalent heat source for the amount of generated
electrical power and take it into account. Finally, this resulted in a
heat source equivalent of 650 W/m2, which roughly corresponds
with an equivalent absorption coefficient of 0.74 (the best corre-
lation of results was achieved by the previously specified magni-
tude of absorption coefficient). In that manner, a realistic result was
gained (a realistic CFD model was developed), which corresponds
well with the experimental model when the PV panel back side is in
question (which was the focus of our paper). Maximal tempera-
tures for each element, during the CFD analysis, are also specified in
Table 4.
The CFD results for Case 1 (Table 4) are depicted in Figs. 13 and
14, while a comparison between measured and simulated values is
presented in Table 4. The characteristic separation zone “A” is
illustrated in Fig. 13 and is accompanied with a temperature in-
crease as shown in Fig. 14.
The numerical simulation of characteristic temperatures corre-
sponds reasonably well with the experimental measurements,
except at the T3 measurement point (TC PVF 3, Fig. 4). The overall
maximum temperature of the individual element showed a better
Fig. 13. Air velocity magnitude distribution of surrounding air on backside surface of
PV panel for Case 1.
agreement with the T3 measurement point than with the tem-
perature at the same location. The possible reason for this deviation
of results was investigated in the variation of Case 1 with a different
wind direction as explained further.
The variation of case 1 (Table 4) with a different angle of attack
was conducted to investigate changes in the temperature field.
Various angles resulted in different temperature distributions on
the back surface of the PV panel, with ‘high’ temperatures always
occurring at locations with high separation zones. For a wind angle
of attack in the order of a ¼ 100 , Fig. 15, it is distinctly noticeable
that separation zone B is accompanied by a high temperature rise in
one corner of the panel. Certain variations of flow conditions (as
illustrated in the case with a wind angle of attack a ¼ 100 ) can be
responsible for the deviation of temperature in T3 or some other
physical phenomenon not taken into account in this simulation.
Nevertheless, simulations showed that with small changes in the
considered input parameters, large variations in temperature dis-
tributions can be observed.
CFD results for the considered case 3 are illustrated in Figs. 16
and 17 while the results are compared to those of the experi-
ments in Table 4; in the case considered there is also a good
agreement between the simulated and experimentally gained
values. As it can be seen in Fig. 17, an air flow separation occurs for
the considered flow regime (case 3, Table 4) which certainly has an
impact on panel performance and will be discussed in detail later.
Finally, we can conclude that simplified CFD simulations are in
agreement with the experimental readings and that the developed
model can be used for further analysis.
The grid independency test for the fluid zone was conducted for
steady state CFD and the results are illustrated in Fig. 18. During the

Table 4
Comparison of measurements and results obtained through CFD analysis.

Conditions Results Temperature at measurement points (K)

Glass Cell Backboard

G1 G3 C1 C2 C3 C4 T1 T3

Case 1: v ¼ 1.5 m/s;a ¼ 45 Exp 316 313 321 323 325 330 321 331
G ¼ 837 W/m2 CFD 318 319 320 320 321 320 320 320
max ¼ 326 max ¼ 326 max ¼ 326
Case 2: v ¼ 2.7 m/s;a ¼ 0 Exp 308 306 314 317 318 322 315 322
G ¼ 837 W/m2 CFD 314 317 315 316 318 316 315 318
max ¼ 317 max ¼ 319 max ¼ 326
Case 3: v ¼ 1.0 m/s;a ¼ 0 , G ¼ 863 W/m2 Exp 321 316 328 329 330 333 328 337
CFD 323 326 324 326 327 324 324 327
max ¼ 326 max ¼ 328 max ¼ 328
222 S. Ni
zetic et al. / Energy 111 (2016) 211e225

Fig. 14. a) PV panel temperature distribution, b) Backboard temperature distribution (Case 1).

Fig. 15. Variation of Case 1 (a ¼ 100 ) a) Backboard temperature distribution b) Separation zone.

Fig. 16. Case 3: a) Panel temperature distribution b) Backboard temperature distribution.

mesh independency test, the mesh in solid regions remained the
same as stated earlier in Table 3, while mesh element sizes were
varied only for the fluid zone. Wind speed was set to 1.0 m/s and
equivalent heat source was 600 W/m2. The results illustrate that
after a mesh size of 2 million elements, the change in average
temperature amounts only 1 K for a mesh up to 5 million elements.
Meanwhile, for smaller meshes with 100 thousand elements, the
difference in average cell temperature amounts 10 K. While
selecting a lower number of elements improves computational
time, a mesh of about 2 million elements is selected as a compro-
mise since the results would not be sufficiently accurate otherwise.
4.3. Temperature panel distribution in relation to wind angle of
attack and solar irradiation intensity: A sensitivity analysis
An additional sensitivity analysis, besides the one presented in
the data of Table 4, was provided to determine the impact on PV
panel temperature distribution with respect to constant wind angle
of attack, intensity of solar irradiation and wind velocity. The re-
sults are presented in Fig. 19, where wind speed (air velocity) was
expressed in m/s, irradiation expressed in W/m2 and finally tem-
perature field expressed in K, as average temperature values for the
PV panel.
 S. Ni
zetic et al. / Energy 111 (2016) 211e225
Fig. 17. Air velocity magnitude distribution of surrounding air over the PV panel for
Case 3.
Fig. 18. Average temperature of solid zones for fluid zone grid independency test.
Fig. 19. Average PV panel temperature with variations of wind velocity and irradiation
intensity (for constant wind angle a ¼ 0 ).
To calculate average PV panel temperature for all the values in
the selected range of influential variables (irradiation, wind velocity
and wind angle), a design of experiments was conducted using 200
randomly generated numerical experiments with variations of all
223
three variables. The experiments are marked in the figures by a
number. Since the figures illustrate average cell temperature for a
constant wind angle, only the experiments with a wind angle close
to a ¼ 0 are marked, as shown in Fig. 19. Neural networks were
used for the approximation of numerical experiments and
extrapolation.
In Fig. 19 it is noticeable that the highest PV panel temperatures
occur for low magnitudes of air velocities and high solar intensity
levels. This was expected since the convective heat rejection from
the PV panel is reduced in this specific case. The highest degrada-
tion of PV panel efficiency occurs in the previously specified case.
However, this typical situation is usually moderate in the majority
of Mediterranean countries; but strong northwest and northeast
winds can occur in summer months from sun rise until sunset
(mistral and meltemi winds) and often have relatively high wind
intensity. The mentioned winds have a favorable effect on PV panel
cooling rate and reduce panel electrical efficiency degradation.
The influence of the wind angle of attack on average PV panel
temperature distribution is presented in Fig. 20, which is to be read
as Fig. 18. It is also noticeable that minimal temperature levels of
the PV panel occur in the case of lowest solar intensity levels, where
temperature minimum occurs when the wind angle of attack is at
around 45 and again at 135 . A similar trend is observable for
maximal PV panel temperature occurring at angles of approxi-
mately 20 and 110 . In paper [28], a study of wind effects on PV
systems was conducted, focusing on the drag and lift coefficients of
PV panels. In their study, the separation of flow was also observed,
but the simulation was conducted only regarding aerodynamics
and the effects of flow on temperature were not considered. It is
interesting to observe that the maximum overturning moments
were observed for wind angles of 45 and 135 in Ref. [28] while in
this study, maximum cooling effects occur for the same angles (i.e.
minimum of average temperature).
5. Discussion
CFD analysis showed a specific air flow regime at the back side of
the PV panel where it can be seen that for specific cases of air flow, a
flow separation occurs at different places on the PV panel back side,
as it was depicted in Figs. 16 and 17. The aforementioned flow
separation causes a temperature increase for specific sections of the
PV panel, which eventually causes an overall increase of PV panel
temperature (that is more emphasized in certain regions of the PV
panel as the position of flow separation depends on wind direc-
tion). That can be seen for measuring point 4 (TC Cell 4 in Fig. 4). It is
only reasonable that the temperature of that point should corre-
spond with the temperature of point 1, because both are equally
away from the frame and it can be concluded that the aluminum
frame dissipates most of the heat. However, since the flow sepa-
ration created an air bubble behind point 4 (Fig. 17, cells at the
bottom of the PV module), the heat flow from the back side was
significantly decreased, when compared with point 1, where a
separation bubble also occurs (Fig. 17) but in a less intensive
manner. It can also be noticed that the experimental values corre-
spond well with the numerical values, as far as the back side is
concerned (Table 4). The difference in front side temperature
(measurement points G1 and G3, Table 4) can be explained with the
fact that the thermocouple at the glass was not insulated from the
surrounding ambient air, hence, the thermocouple sensor was
reading the mean temperature of the glass-air border layer. The
previous issue did not occur on the back side of the PV panel, where
the sensors were isolated from the ambient air by tape with similar
properties like the PVF layer. That is why the numerically gained
back side temperatures correspond better with the experimental
values. Further, the numerically gained temperatures of the PV cells
224 S. Ni
zetic et al. / Energy 111 (2016) 211e225
Fig. 20. Average solar cell temperature with variations in wind angle of attack and insolation intensity for constant wind velocity of 1.0 m/s.
seem to correspond well with the experimental model (Table 4) so
the developed CFD model was used for an additional sensitivity
analysis. The sensitivity analysis on the average PV panel temper-
ature field was obtained for different wind angles of attack, wind
velocity and also for different intensities of solar irradiation. From
the gained results, it is noticeable that the highest degradation of
PV panel electrical efficiency occurs in the circumstances of lowest
wind velocity values, highest solar irradiation levels and finally for
the wind angle of attack at around 20 .
For the same irradiation levels and wind velocity, the average PV
panel temperature has a variation of 3  C from maximal to minimal.
The previous results are also useful data for the mounting of PV
panels in Mediterranean climate conditions.
6. Conclusions
An experimental investigation of two commonly used photo-
voltaic technologies (poly-Si and mono-Si) was presented in the
paper, where the main focus was to examine the effect of air
backside convective thermal profile, i.e. its potential impact on PV
panel performance. The experimental measurements were carried
out in typical Mediterranean climate operating conditions from
April to July. Besides an experimental approach, a CFD analysis was
also obtained as well as a sensitivity analysis. The developed CDF
model lead to results that compared well to the experimental
measurements; the CFD model generally provided slightly higher
values for specific measurement points compared to the experi-
mental readings (higher only for glass and generally lower for all
other measurement points). The approach that was used in the CFD
modeling was related to the introduction of an equivalent heat
source, where we managed to take the influence of thermal radi-
ation in an indirect manner. Finally, because of the previous reason,
we succeeded in achieving a good agreement between the
measured and numerical results.
The sensitivity analysis was carried out, with respect to the
relation of PV panel temperature distribution and the parameters of
wind velocity, wind angle of attack and solar insolation intensity.
Furthermore, the detected specific range of the considered influ-
ential parameters leads to the highest PV panel electrical efficiency
degradation rate. In general, the most critical period for PV panel
efficiency degradation is in regimes of highest solar irradiation
levels and lowest wind air velocities.
The most interesting finding of the study is the detection of two
typical and occurring convective air thermal profiles at the backside
of the PV panel for typical Mediterranean climate operating con-
ditions. Regarding the detected convective thermal air profiles,
there is no significant difference between mono and poly-
crystalline PV technology. According to the measurements ob-
tained, the most important discovered effect is flow separation that
occurs at the PV panel's backside, which in essence acts as a kind of
thermal insulation and which also causes a specific shape of
backside convective thermal profile. The detected effect of flow
separation results in the increase of average PV panel temperature
and finally enhances the degradation rate of panel electrical effi-
ciency. According to the available experimental readings, the
average increase in PV panel temperature due to the discovered
effect of flow separation ranges from 5  C up to 9  C, which means
that the estimated degradation of PV panel electrical efficiency can
range from 2.5% up to 4.5% in average (if flow separation would be
removed, an increase in efficiency would surely be provided).
From the previous elaboration, an important conclusion could
be drawn and it is certainly related to the consideration of possibly
redesigning existing commonly used PV panels in order to break
the addressed flow separation. The PV panel redesign could be
simple and of rather limited initial costs. The obtained modification
of PV panel backside surface could lead to an average electrical
efficiency increase.
Further research is needed to determine the aerodynamics of
the redesign and also to evaluate other profiles of air convective
thermal profile that can possibly occur due to stochastic sur-
rounding conditions.
References
[1] Singh GK. Solar power generation by PV (photovoltaic) technology: a review.
Energy 2013;53:1e13.
[2] Tiwari GN, Mishra RK, Solanki SC. Photovoltaic modules and their applica-
tions: a review on thermal modelling. Appl Energy 2011;88(7):2287e304.
[3] Web source: https://www.irena.org/DocumentDownloads/Publications/
IRENA-ETSAP%20Tech%20Brief%20E11%20Solar%20PV.pdf, Last visited,
November 10, (2015).
[4] Alami AH. Effects of evaporative cooling on efficiency of photovoltaic modules.
Energy Convers Manag 2014;77:668e79.



[5] Ni
zeti
c S, Coko D, Yadav A. Grubi
sic-Cabo. Water spray cooling technique
applied on a photovoltaic panel: the performance response. Energy Convers
Manag 2016;108:287e96.
[6] Bahaidarah H, Subhan A, Gandhidasan P, Rehman S. Performance evaluation of
a PV (photovoltaic) module by back surface water cooling for hot climatic
conditions. Energy 2013;59:445e53.
[7] Valeh-E-Sheyda P, Rahimi M, Parsamodghadam A, Masahi MM. Using a wind
 S. Ni
zetic et al. / Energy 111 (2016) 211e225 225

driven ventilator to enhance a photovoltaic cell power generation. Energy thermal energy recovery. J Renew Sustain Energy 2014;6. http://dx.doi.org/
Build 2014;73:115e9. 10.1063/1.4885178.
[8] Teo HG, Lee PS, Hawlader MNA. An active cooling system for photovoltaic [29] Jubayer CM, Hangan H. Numerical simulation of wind effects on a stand-alone
modules. Appl Energy 2012;90(1):309e15. ground mounted photo voltaic (PV) system. J Wind Eng Ind Aerodyn
[9] Smith CJ, Forster PM, Crook R. Global analysis of photovoltaic energy output 2014;134:56e64.
enhanced by phase change material cooling. Appl Energy 2014;126(1):21e8. [30] Vaillon R, Robin L, Muresan C, Me nezo C. Modeling of coupled spectral radi-
[10] Chandrasekar M, Suresh S, Senthilkumar T, Ganesh M. Passive cooling of ation, thermal and carrier transport in a silicon photovoltaic cell”. Int J Heat
standalone flat PV module with cotton wick structures. Energy Convers Mass Transf 2006;46:4454e68.
Manag 2013;71:43e50. [31] Allegrini J, Orehounig K, Mavromatidis G, Ruesch F, Dorer V, Evins R. A review
[11] Tsai H-L. Modeling and validation of refrigerant-based PVT-assisted heat of modelling approaches and tools for the simulation of district-scale energy
pump water heating (PVTA-HPWH) system. Sol Energy 2015;122:36e47. systems. Renew Sustain Energy Rev 2015;52:1391e404.
[12] Elsafi AM, Gandhidasan P. Comparative study of double-pass flat and com- [32] Menter FR. Two-equation eddy-viscosity turbulence models for engineering
pound parabolic concentrated photovoltaic-thermal systems with and applications. AIAA J 1994;32(8):269e89.
without fins. Energy Convers Manag 2015;98:59e68. [33] Defraeye T, Blocken B, Carmeliet J. CFD analysis of convective heat transfer at
[13] Rejeb O, Dhaou H, Jemni A. A numerical investigation of a photovoltaic the surfaces of a cube immersed in a turbulent boundary layer. Int J Heat Mass
thermal (PV/T) collector. Renew Energy 2015;77:43e50. Transf 2010;53(1e3):297e308.
[14] Liang R, Zhang J, Ma L, Li Y. Performance evaluation of new type hybrid
photovoltaic/thermal solar collector by experimental study. Appl Therm Eng
2015;75:487e92.
Nomenclature
[15] Tourkov K, Schaefer, L.. Performance evaluation of a PVT/ORC (photovoltaic
thermal/organic Rankine cycle) system with optimization of the ORC and a: overall coefficient of absorptivity
evaluation of several PV (photovoltaic) materials. Energy 2015;82:839e49. GS: solar irradiance (insolation), W/m2
[16] Zafar S, Dincer I. Thermodynamic analysis of a combined PV/T-fuel cell system E_ in : overall energy input into the PV panel, W
for power, heat, fresh water and hydrogen production. Int J Hydrogen Energy E_ out : overall energy output from the PV panel, W
2014;39(19). E_ current : electric output from the PV panel, W
[17] Zhao J, Zhou X, Ma Y, Liu W. A novel maximum power point tracking strategy Q_ cell : PV cell heat source, W
based on optimal voltage control for photovoltaic systems under variable Q_ in : energy (heat) income due to solar insolation, W
environmental conditions. Sol Energy 2015;122:630e49. Q_ convection : convection heat flow, W
[18] Li S. A maximum power point tracking method with variable weather pa- Q_ radiation : thermal radiation heat flow, W
rameters based on input resistance for photovoltaic system. Energy Convers Ap: Surface of the PV panel, m2
Manag 2015;106:290e9. T: thermodynamic temperature, K
[19] Ahmed J, Salam Z. An improved perturb and observe (P&O) maximum power Tamb: temperature of the surrounding objects, K
point tracking (MPPT) algorithm for higher efficiency. Appl Energy 2015;150: p: pressure, Pa
97e108. t: time, s
[20] Chao KH. An extension theory-based maximum power tracker using a particle ui : mean flow Cartesian velocity components
swarm optimization algorithm. Energy Convers Manag 2014;86:435e42. xj: Cartesian coordinates
[21] Abu Eldahab YE, Saad NH, Zekry A. Enhancing the maximum power point fi: body forces
tracking techniques for photovoltaic systems. Renew Sustain Energy Rev Rij: Reynolds stress tensor
2014;40:505e14. k: turbulence kinetic energy
[22] Daraban S, Petreus D, Morel C. A novel MPPT (maximum power point h: sensible enthalpy, J/kg
tracking) algorithm based on a modified genetic algorithm specialized on kc: thermal conductivity, W/mK
tracking the global maximum power point in photovoltaic systems affected by kt: thermal conductivity due to turbulence, W/mK
partial shading. Energy 2014;74:374e8. Sh: volumetric heat source, W/m3
[23] Feng S, Cao L, Fang G. Dynamic performances modeling of a photo-
voltaicethermal collector with water heating in buildings. Energy Build Greek symbols:
2013;66:485e94.
[24] Zhu T, Ertekin E. Phonon transport on two-dimensional graphene/boron
nitride superlattices. Phys Rev B 2014;90.19:195209. hcell: electrical efficiency of PV cell,
[25] Matic Z. Solar radiation atlas for the area of the republic of Croatia. first ed. DTeq.: temperature difference between PV panel surface temperature and sur-
Zagreb: Energy Institute Hrvoje Pozar; 2007. rounding air, K
[26] Menter FR. Zonal two equation k-u turbulence models for aerodynamic flows. a: average heat transfer coefficient, W/m2K
1992. NASA technical memorandum 103975. ε: emissivity factor,
[27] Brinkworh BJ, Cross BM, Marshall RH, Yang H. Thermal regulation of photo- s: Stefan-Boltzmann constant, W/m2K4
voltaic cladding. Sol Energy 1997;61(3):169e78. r: density, kg/m3
[28] Hiren D. Computational fluid dynamics analysis and experimental validation m: molecular viscosity,
of improvement in overall energy efficiency of a solar photovoltaic panel by mt: turbulent viscosity.