ARTICLE IN PRESS

Solar Energy Materials & Solar Cells 92 (2008) 1393– 1402

Contents lists available at ScienceDirect

Solar Energy Materials & Solar Cells

journal homepage: www.elsevier.com/locate/solmat

A simple correlation for the operating temperature of photovoltaic modules

of arbitrary mounting
E. Skoplaki, A.G. Boudouvis, J.A. Palyvos 
Solar Engineering Unit, School of Chemical Engineering, NTUA, Athens GR-15780, Greece

a r t i c l e in fo abstract

Article history: Following a brief discussion regarding the operating temperature of commercial grade silicon
Received 24 July 2007 photovoltaic (PV) cells/modules and its effect upon the performance of free-standing one-sun PV
Accepted 27 May 2008 installations, a simple semi-empirical explicit correlation for PV cell temperature and the corresponding
Available online 10 July 2008
efficiency form are proposed for modules of arbitrary mounting. To this end, a dimensionless mounting
Keywords: parameter, o, is introduced rendering the correlations suitable for systems like building-integrated
PV cell operating temperature photovoltaic (BIPV) array generators. The implications of ignoring radiation and free-convection are
Solar cell quantified and a comparison is made with analogous relations in the literature.
PV cell efficiency & 2008 Elsevier B.V. All rights reserved.
Temperature dependence of PV efficiency
Mounting parameter

1. Introduction From the mathematical point of view, the correlations for the
PV operating temperature are either explicit in form, i.e. they give
It is well established that temperature plays a central role in Tc directly, or they are implicit, that is they involve variables which
the photovoltaic (PV) conversion process since it affects basic themselves depend on Tc. In this last case, an iteration procedure is
electrical quantities, such as the voltage and the current of the PV necessary for the relevant calculation. Most of the correlations
generator. As a consequence, the operating temperature, Tc, of a PV usually include a reference state and the values of the pertinent
device, whether a simple module, a PV/Thermal collector or a variables at this reference state. It goes without saying that the
building-integrated photovoltaic (BIPV) array, represents a funda- degree of accuracy with which Tc can be estimated, has a direct
mental variable which directly affects the electrical power output bearing on any simulation results and/or relevant calculations
of the device and its efficiency. It is only natural, therefore, that which require this temperature as an input.
in recent years, its prediction and its influence on module/
array operation has received considerable attention among the
2. PV cell/module efficiency and power as a function of the
scientific community.
operating temperature
The proposed correlations in the literature normally express Tc
as a function of the pertinent weather variables, namely ambient
temperature, Ta, and local wind speed, Vw, as well as of the solar The operating temperature effect on the electrical power
radiation flux (or irradiance) incident on the plane of the array, GT. produced by a PV cell or module can be traced to the
Generally speaking, Tc is extremely sensitive to wind speed, less so temperature’s influence upon the current, I, and the voltage, V,
to wind direction, and practically insensitive to the atmospheric as the maximum power is given by
temperature [1]. On the other hand, it obviously depends strongly Pm ¼ V m Im ¼ ðFFÞV oc Isc (1)
on the ‘‘plane-of-array’’ irradiance. These correlations also
In this expression, which also serves as a definition of the fill
include, as parameters, material and system-dependent proper-
factor,1 FF, subscript m refers to the maximum power point in the
ties such as glazing-cover transmittance, t, plate absorptance, a,
module’s I–V curve, while subscripts oc and sc denote open circuit
etc. A large number of correlations can also be found in the
and short circuit values, respectively. It turns out that both the
literature, which express the negative effect that an operating
temperature increase has upon the electrical efficiency, Zc, of a PV
cell/module/array. 1
The fill factor is a measure of how much series resistance and how little
shunt resistance there is in a solar cell and its circuit. It affects the maximum
power voltage change with irradiance and decreases with module age, as the series
 Corresponding author. Tel.: +30210 7723297; fax: +30210 7723298. resistance increases in a degrading PV cell ([2], p. 85). For silicon solar cells, a
E-mail address: jpalyvos@chemeng.ntua.gr (J.A. Palyvos). healthy Fill Factor lies in the range 0.75–0.85.

0927-0248/$ - see front matter & 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.solmat.2008.05.016
 ARTICLE IN PRESS

1394 E. Skoplaki et al. / Solar Energy Materials & Solar Cells 92 (2008) 1393–1402

Nomenclature b(GT) Voc correction coefficient for temperature (1C1)

bref efficiency correction coefficient for temperature
A cell/module aperture surface area, m2 (1C1)
FF fill factor g efficiency correction coefficient for irradiance
GT irradiance (solar rad’n flux) on module plane, W/m2 d(T) Voc correction coefficient for irradiance
hfree free-convection heat transfer coefficient, W/m2K D ¼ Tc–Ta
hrs radiation heat transfer coefficient, W/m2K eg PV glazing emmissivity
hw wind-convection heat transfer coefficient, W/m2K Zc cell/module electrical efficiency
I electrical current ZT ref electrical efficiency at temperature Tref
k Ross coefficient—Eq. (14), Km2/W s Stefan–Boltzmann constant (5.6697  108 W/m2K4)
INOCT installed nominal operating cell temperature, 1C t transmittance of glazing
NOCT nominal operating cell temperature, 1C o mounting coefficient—Eq. (24)
NTE nominal terrestrial environment (cf. SRE)
P electrical power, W Subscripts
SOC standard operating conditions
SRC standard reporting conditions a ambient
SRE standard reference environment (cf. NTE) b back side
TRE test reference environment c cell/module
Ta ambient temperature, K f free-stream
Tb back-side cell temperature, K L thermal loss
Tc cell/module operating temperature, K m at maximum power point
Ts sky temperature, K NOCT at NOCT conditions
UL overall thermal loss coefficient, W/m2K NTE at NTE conditions
V voltage, V oc open circuit
Vw wind speed, m/s ref at reference conditions
s sky
Greek symbols sc short circuit
T on module’s tilted plane
a solar absorptance of PV layer w wind induced
a Isc correction coefficient for temperature (1C1)

open circuit voltage and the fill factor decrease substantially with irradiance, d(T) is the Voc correction coefficient for solar radiation
temperature (as the thermally excited electrons begin to dominate flux as a function of the module temperature, and the subscript
the electrical properties of the semi-conductor), while the short- zero refers to SRC. If Eqs. (2) and (3) are introduced into Eq. (1)
circuit current increases slightly [3]. and combined with the definition of the efficiency, they can
The same argument holds for the effect of temperature on the produce an expression for the latter. In view of the fact that
electrical efficiency of the PV cell/module, which is defined as the correction coefficients are small (a is of order 104, b is of
Zc ¼ Pm/AG, with A the cell’s area or the module’s aperture area. For order 103, and d is of order 102), products of coefficients can
this definition to hold, the maximum power must be measured be dropped, and after a little algebra a linear relation for Zc
under the so-called Standard Reporting Conditions2 (SRC). emerges, similar in form with the well known equation proposed
Equations for the short-circuit current and the open-circuit voltage by Evans [8]:
as functions of the PV operating temperature can be found in the
literature, based on various proposed electrical models (cf. [5] and Zc ¼ ZT ref ½1  bref ðT c  T ref Þ þ glog10 GT  (4)
references therein). Such expressions include temperature coefficients
that provide the rate of change with respect to temperature of PV
performance parameters such as Isc, Voc, Imp, Vmp, but also FF, and Zc. In this expression, ZT ref is the module’s electrical efficiency at
Recent standard testing methods [4] address two of these parameters, temperature Tref and at solar radiation flux of 1000 W/m2. The
one for current and one for voltage, although outdoor characterization efficiency correction coefficient for temperature, bref, and the
of module performance studies have suggested that all four current efficiency correction coefficient for solar irradiance, g, are mainly
and voltage parameters listed above are needed for accurate material properties, having average values of about 0.0045 K1
predictions under various operating conditions [6]. and 0.12, respectively, for silicon modules [9]. The latter, however,
Among the relevant expressions, those used in the NREL’s is usually taken as zero [8], so that Eq. (4) reduces to
Module Energy Rating (MER) procedure [7] are
Zc ¼ ZTref ½1  bref ðT c  T ref Þ (5)
GT
SC ¼ ISC ½1 þ aðT c  T 0 Þ (2)
GT 0 0
which represents the traditional simple linear expression for the
and
PV electrical efficiency [10].
V OC ¼ V OC0 ½1 þ bðGT 0 ÞðT c  T 0 Þ½1 þ dðT c Þ lnðGT =GT 0 Þ (3) The quantities ZT ref and bref are usually given by the PV module
manufacturers, but they can be also obtained from flash tests, in
in which a is the Isc correction coefficient for temperature, b(GT) is
which the module’s electrical output is measured at two different
the Voc correction coefficient for temperature as a function of
temperatures for a given solar radiation flux [11]. The value of the
efficiency correction coefficient for temperature, on the other
2
1000 W/m2 total irradiance, G 159 spectral irradiance, 25 1C [4] hand, depends not only on the PV material but also on Tref. It is
 ARTICLE IN PRESS

E. Skoplaki et al. / Solar Energy Materials & Solar Cells 92 (2008) 1393–1402 1395

given by the following expression: mounting frame oriented ‘‘normal to solar noon’’ [17,14]. In symbols3
1 NOCT ¼ ðT c  T a ÞNTE þ 20ð CÞ (8)
bref ¼ (6)
T Z¼0  T ref
The relevant method, which assumes that both sides of the
in which TZ ¼ 0 is the (high) temperature at which the PV
module ‘‘feel’’ the same ambient temperature, is based on the
module’s electrical efficiency drops to zero [12] (this temperature
observation that the temperature difference (Tc–Ta) is practically
is 270 1C for common silicon solar cells [13]). For the usual
independent of Ta and just linearly proportional to the incident solar
reference temperature Tref ¼ 25 1C, a literature average value for
radiation flux, especially at irradiance levels above 400 W/m2 [21].
crystalline silicon modules reference efficiency is ZT ref E0.12.
In addition, it makes the approximation that the overall thermal loss
coefficient, UL, is constant, to be obtained by the NOCT test [22]. This
3. Operating temperature prediction for free-standing PV approximation neglects the effect upon UL of such factors as wind
arrays speed, humidity, and temperature, even though they can substan-
tially affect it, because the effect on the resulting absolute operating
The thermal environment which establishes the instantaneous temperature of the cell is relatively small [23].
value of the PV module’s operating temperature is quite complex. If the steady-state energy balance, namely
As a result of internal processes taking place within the [PV electrical power] ¼ [absorbed solar power]–[dissipated
semiconductor material during its exposure to photons, a very power]
large portion of the incident solar radiation flux is degraded and is written for a unit module area, i.e.
released as heat. Thus, standard heat transfer mechanisms, such
as convection and radiation, must be taken into account in the Zc GT ¼ ðtaÞGT  U L ðT c  T a Þ (9)
appropriate energy balance on the module, i.e. in the procedure
a simple rearrangement leads to the following expression for the
which leads to the prediction of Tc. In most cases these
operating temperature [23]:
mechanisms affect both the front and the back side of the module
as, in typical installations, provision is usually made to facilitate   
GT Z
the removal of the rejected heat so that the module can operate as T c ¼ T a þ ðtaÞ 1 c (10)
UL ðtaÞ
efficiently as possible. In the case of rack-mounted/free-standing
arrays, heat conduction through the mounting frame should This last equation can be applied to NOCT conditions (i.e. for
normally also be taken into account. However, at steady-state Zc ¼ 0) and solved for the assumed constant UL, which is found to
conditions, conduction merely transports heat toward the sur- be
faces which release it to the atmosphere via convection and
radiation ([14], p. 240, ch. 8). ðt aÞGNOCT
UL ¼ (11)
The traditional steady-state energy balance which leads to the T NOCT  T a;NOCT
prediction of Tc requires as input
Substitution of this last expression back into Eq. (10), then, leads
to the equation
 Thermal and physical properties of the cell/module.
 The solar resource and weather data.    
GT Z
 The wind heat transfer coefficient. Tc ¼ Ta þ ðT NOCT  T a;NOCT Þ 1  c (12)
GNOCT ðtaÞ

This last quantity is not easy to determine, especially in the where for simplicity, we have written GNOCT and TNOCT instead of
field, where monitoring the wind or establishing uniform GT,NOCT and Tc,NOCT, respectively [4,24].
conditions for the necessary measurements is not an easy job. The constant UL approximation, however, is not really
This explains the large number of wind loss coefficient correla- necessary, as it does not lead to a substantial simplification. Thus,
tions which have appeared in the literature (cf. Ref. [15]). if the energy balance equation (9) is written for NOCT conditions
The temperature of the individual cells (actually, that of (i.e. with Zc ¼ 0) and the resulting equation is combined with
their p–n junctions) within a PV module, i.e. Tc, is the proper Eq. (9), an expression for Tc similar to Eq. (12) can be easily
temperature to use in order to predict the electrical performance obtained in the form ([25], Section 23.3)
of the module. However, it may be higher by a few degrees from     
the back-side temperature, Tb, and this difference depends mainly GT U L;NOCT Z
Tc ¼ Ta þ ðT NOCT  T a;NOCT Þ 1  c (13)
on the materials of the module substrate and on the intensity of GNOCT UL ðtaÞ
the solar radiation flux. The relation between the two tempera-
But since Zc is itself a function of Tc, as was shown in Section 2,
tures can be expressed by the simple linear expression
Eq. (13) is an implicit equation for the PV cell/module operating
temperature, suitable mainly for situations in which the modules
GT
Tc ¼ Tb þ DT Gref (7) are mounted in the free standing manner for which the NOCT
Gref
methodology holds. That is, one should avoid using it for BIPV
installations, where the two sides of the modules are subjected to
in which Gref is a reference solar radiation flux incident on the quite different environmental conditions and, thus, require
module (1000 W/m2), and DT Gref is the temperature difference modified prediction approaches (The NOCT model here can
between the PV cells and the module back side at this reference under-predict Tc by as much as 20 K [24]).
solar radiation flux [16].
An established procedure to formulate Tc via the energy balance
3
on the cell/module involves use of the so-called nominal operating This quantity is determined from actual measurements of cell temperature
cell temperature (NOCT), defined as the temperature of a device at for a range of environmental conditions similar to the NTE and reported by the
module’s manufacturer [18]. When a module operates at NOCT, the term ‘‘standard
the conditions of the nominal terrestrial environment (NTE): Solar operating conditions’’ (SOC)is sometimes used. In an alternate definition the
irradiance 800 W/m2, ambient temperature 20 1C, average wind number 4572 1C is used in place of NOCT [19]. Also, the term standard reference
speed 1 m/s, zero electrical load (i.e. open circuit), and free-standing environment (SRE) is recently used to denote NTE [20].
 ARTICLE IN PRESS

1396 E. Skoplaki et al. / Solar Energy Materials & Solar Cells 92 (2008) 1393–1402

4. Methodological approaches—extensions to BIPV situations on the respective layer and featuring the respective temperature,
i.e. that of the PV module, of the air in the gap, and of the
The open rack-mounting condition, i.e. fixing the modules on a wall. Such balances consider all the applicable modes of heat
free-standing frame open to the ambient, which is a basic element transfer among the layers, the ambient, and (occasionally) the
of the NOCT methodology, seems at first prohibitive for the use of interior space and, clearly, can vary widely in detail. In most
NOCT-based correlations in the BIPV modeling/design area, cases, the methodology adopted involves a lumped analysis
because of the aforementioned possible error in the Tc value. approach, typically assuming uniform conditions across the
The ever increasing interest in BIPV applications, however, gap, i.e. bulk flow, while, in more detailed studies, dynamic
brought forward the need for a proper estimation of NOCT, which models and CFD methods are employed to provide additional
would take into account the integration-dependent deviation spatial detail [32–34].
from NTE/SRE conditions. That is, the proper NOCT to use should
accommodate the BIPV encountered angles of incidence of say
901, and not just 01, as well as the observed higher module-back 5. Explicit correlations for the PV operating temperature
temperatures, due to lack of proper cooling from the poorly
ventilated back side.4 Thus, the module’s correct NOCT, which The simplest explicit equation for the steady-state operating
obviously depends on the mounting scheme for a given solar temperature of a solar cell/module, links Tc with the ambient
radiation flux level, must be measured in a properly designed temperature and the incident solar radiation flux:
and well controlled outdoor test bed, like the BIPV test facility
T c ¼ T a þ kGT (14)
at the US National Institute of Standards and Technology [24] or
the European Commission’s Test Reference Environment (TRE) In this linear expression, which holds for no load and no wind,
rig, that was recently set up at the JRC Ispra specially for BIPV the constant k, known as the Ross coefficient, expresses the
testing [20]. temperature rise above ambient with increasing solar flux, i.e.
In the spirit of the above, the US Sandia National Laboratory k ¼ D(TcTa)/DGT) [35]. The main difficulty with this linear model
developed a simple PV thermal model for the prediction of the lies in the estimation of k, which can be measured for an installed
operating temperature, in which the major input parameter is the array but not easily estimated beforehand, especially when the
00
Installed00 Nominal Operating Cell Temperature (INOCT). It is wind effects are important.5 Earlier reported values for k were in
defined as the PV cell temperature of an installed array at the range 0.02–0.04 1C m2/W [2,36], but a more recent IEA study
NOCT conditions [27]. The model was incorporated into the extended this range upwards, categorizing the results qualita-
PVFORM performance simulator, which uses INOCT to character- tively according to level of integration and size of air-gap (if any)
ize the thermal properties of the module and its mounting behind the modules [37]. Table 1 lists values of parameter k for
configuration. INOCT can be estimated from the manufacturer’s various array types/mounting schemes, estimated from the plots
NOCT value and a set of empirically determined rules regarding shown in the above study.
the mounting configuration [28], or from cell temperature data of A similar explicit equation can be derived through simplifica-
a fielded array. In the estimation, the thermal capacitance of the tion of Eq. (13). If the usual assumption of constant UL is made,
module is also included, to account for the natural temperature then Eq. (13) can be written as ([38], p. 88)
lag [29]. A typical value for rack-mounted PV modules is
 
INOCT ¼ NOCT3 1C [30]. T NOCT  T a;NOCT
Tc ¼ Ta þ GT (15)
An analogous European PV energy rating effort at Ispra, GNOCT
bypassed the restrictions of NOCT by estimating the Nominal
Operating Specific Temperature or NOST, defined as the site and since the term (Zc/ta) is small compared to unity (cf. [25], p. 760).
mounting specific module temperature of a module operating at This simple equation obviously expresses the same linear
the maximum power point, under 800 W/m2 irradiance and 20 1C dependence of Tc upon GT as in Eq. (14), and it only produces a
ambient temperature. The value of NOST is established from the 2–3 1C error in the calculated Tc, at a typical average solar flux of
complete measured data set by plotting the module and ambient 600 W/m2 and for an assumed ta value of 0.9, regardless of the
temperature differences against irradiance, just like in the value of Zc [22].
standard NOCT determination procedure, with the transformation The realistic assumption that the electrical efficiency is a
on the measured data following a quite similar equation: linear function of temperature can also lead to explicit forms for
Tc ¼ (NOST201)GT/800+Ta. A sensitivity analysis showed an the operating temperature of solar cells/modules. Combination of
energy prediction variation of 1.5% per 5 1C increase in the Eq. (13) with Eq. (5) gives
NOST, implying that the overall error in energy prediction will be
   
small if approximate NOST values are used [31]. (The same GT U L;NOCT
Ta þ GNOCT ULðT NOCT  T a;NOCT Þ 1  Zta ref
ð1 þ bref T ref Þ
argument can be used for NOCT itself, where errors of about 73 1C Tc ¼   
GT U L;NOCT
in NOCT value introduce minor error—about 71.5%—on annual 1  breftaZref GNOCT UL ðT NOCT  T a;NOCT Þ
performance calculations, since the temperature has a second
(16)
order influence on module energy output [26].)
Actually, in modern BIPV situations, the PV modules are
which is still implicit in Tc through the radiation component of UL.
typically mounted at an optimized distance from the building
It turns out, however, that in many situations with flat plates
fac- ade and, thus, the energy balance is also written for the air
exposed to the ambient, the radiation losses are 3–4 times
layer in the gap between module and wall, as well as for the wall
smaller than the wind induced convection losses [39]. Therefore,
itself. That is, Eq. (9) is replaced by a system of three simultaneous
radiation (and free convection, which is very small at the usual
equations, each one resulting from an individual energy balance
wind speeds) can be ignored, in which case the ratio UL,NOCT/UL
can be replaced by the ratio of the wind convection coefficients,
4
Thermal inertia is another problem in the NOCT calculation, as Tc– Ta is
5
higher in the afternoon for the same GT values. This can cause a 73 1C error in A wind dependence of k has been introduced in the form
NOCT but, still, only a 71.5% error in annual performance estimations [26]. k(Vw) ¼ c1+c2exp(c3Vw) [11].
 ARTICLE IN PRESS

E. Skoplaki et al. / Solar Energy Materials & Solar Cells 92 (2008) 1393–1402 1397

Table 1 The relative sizes of the two major modes of heat losses, then,
Values of the Ross parameter k (Eq. (14)) for various mounting situationsa are given in Fig. 2, in the form of the ratio hw/hrs, for the
‘‘bounding’’ environmental conditions that hold for the above
PV array mounting type k (Km2/W)
ranges of ambient temperature and irradiance. It can be seen that
Free standing 0.021 this ratio goes down to about 1.0 at Vf ¼ 1 m/s, suggesting that at
Flat roof 0.026 this wind speed the radiation coefficient is comparable to the
Sloped roof: Well cooled 0.020 wind coefficient, exceeding the latter at wind speeds less than
Sloped roof: not so well cooled 0.034
1 m/s, at low irradiance. At the other end of wind speeds, as
Sloped roof: highly integrated, poorly ventilated 0.056
Fac- ade integrated: transparent PV’s 0.046 expected, wind convection dominates, since the ratio can reach
Fac- ade integrated: opaque PVs—narrow gap 0.054 3.3 at Vf ¼ 15 m/s. These ‘‘bounding’’ values are listed in Table 2.
In spite of the obviously non-negligible size of the radiation
a
Adapted from data in Ref. [37]. component, however, its inclusion (and that of free convection, as
well) in the calculation of the PV operating temperature using
Eq. (17) has only a minor effect, compared to the corresponding
hw,NOCT/hw, leading to the explicit form: calculation of Tc using only the wind convection component. Such
a comparison is shown in Fig. 3 for the ‘‘bounding’’ environmental
  h i
Ta þ GT hw;NOCT
ðT NOCT  T a;NOCT Þ 1  ðtaÞZref
ð1 þ bref T ref Þ conditions only, since the corresponding curves for all the other
GNOCT hw
Tc ¼    (17) combinations of Ta and GT fall in-between the depicted curves.
GT hw;NOCT
1  brefðtaÞ
Zref
GNOCT hw
ðT NOCT  T a;NOCT Þ In the Vf range 1–15 m/s the Tc value calculated on the basis of
wind convection only is always lower by less than 4 1C from the
This last equation, represents the most general expression for corresponding value which includes radiation and free convection
the operating temperature if free convection and radiation losses in the calculation. In fact, this difference, when averaged over the
to the ambient can be ignored. The implications of this wind speed range 1–15 m/s for the given Ta and GT combination,
simplification are discussed in the section that follows. lies in the range 1.3–3.6 1C, as shown also in Table 2.
At Vf ¼ 0, the simple calculation for Tc, i.e. the one ignoring
radiation and free convection, produces values that are higher by
2.4–3.2 1C compared to the more accurate calculation values. In
6. Free convection and radiation contributions to the
the range 0.2o Vfo0.8m/s the difference is in the range
operating temperature
1.5–1.7 1C, and at Vf ¼ 1 m/s the two calculations produce
practically identical results at high irradiance and a maximum
In the development of Eq. (17), both free convection and
difference of 0.9 1C at 600 W/m2. Thus, in view of the fact that a
radiation were ignored as being small compared to forced
wind speed of 2 m/s represents ‘‘normal conditions’’ [46], it can be
convection, an assumption which is not realistic at very small
safely assumed that (with the exception of zero wind speed) the
wind speeds, i.e. speeds approaching zero.
calculation of Tc can be based on wind convection only and that
The sizes of the three heat transfer coefficients that comprise
the expected error of ignoring radiation is less than 2 1C.
UL, are shown in Fig. 1, together with the resulting values of Tc as
predicted by Eq. (17), for the range of wind speeds 0–15 m/s, at
Ta ¼ 30 1C and GT ¼ 1000 W/m2. The individual coefficients have
7. Proposed simple correlations for PV operating temperature,
been calculated using standard equations: For radiation (referred
efficiency, and electrical power output
to Ta)6 the linearized form [41]:

hrs ¼ sg ðT 2c þ T 2s ÞðT c þ T s ÞðT c  T s Þ=ðT c  T a Þ (18) A close inspection of Eq. (17) reveals that it can be further
simplified. If values for the ‘NOCT variables’ and average reference
for free convection the simple approximation7 ([44], p. 229). quantities (cf. Section 3) are substituted into Eq. (17), the second
term in the denominator is of order 0.044 at GT ¼ 1000 W/m2 or
hfree ¼ 1:78ðT c  T a Þ1=3 (19)
0.027 at GT ¼ 600 W/m2, i.e. much smaller than 1 and, therefore,
and for forced (wind) convection the purely empirical linear can be omitted. Moreover, according to Eq. (20), hw,NOCT ¼ 8.91+
correlation [45] 2.0(1) ¼ 10.91 W/m2K (since, by definition, Vf,NOTC ¼ 1 m/s), while
GNOTC ¼ 800 W/m2, and Ta,NOCT ¼ 20 1C. On the other hand, typical
hw ¼ 8:91 þ 2:0V f (20)
manufacturer’s specifications for p-Si PV modules give
In the above equations (18)–(20), s is the Stefan–Boltzmann Tc,NOCT ¼ 47 1C72 1C, ZrefE0.12 and bref ¼ 0.004 1C1 for Tref ¼ 25 1
constant, eg is the PV glazing emissivity, Ts is the sky temperature, C. Thus, for the reasonable value ta ¼ 0.9, Eq. (17) reduces to the
and Vf is the free stream wind speed in the windward side of the following simple semi-empirical equation:
PV array [45]. Fig. 1, which represents an upper bound of the  
0:32
radiation and free convection contribution to Tc for ambient Tc ¼ Ta þ GT ðV f 40Þ (21)
8:91 þ 2:0V f
temperatures down to 10 1C (Ts 10 1C), clearly shows that free
convection is very small over the wind speed range 115 m/s. which correlates the PV module operating temperature with the
Radiation, however, is comparable to the otherwise dominant three basic environmental variables and which applies to PV arrays
wind convection component, when the wind speed approaches mounted on free-standing frames, for wind speeds 40 m/s. The
zero. The exactly same picture emerges for all the combinations of predictions of Eq. (21) are compared in Fig. 4 with ‘‘average over V00 f
irradiance and ambient temperature in the ranges 600o GTo1000 experimental data from the extensive IEA PVPS database [37]. The
W/m2 and 10o Tao30 1C. plots give the operating temperature rise over the ambient,
DT ¼ Tc– Ta, as a function of irradiance and wind speed. The dashed
6
curve in Fig. 4 represents the trend-line in the relevant experimental
Unlike convection, the natural ‘‘heat sink’’ for radiation is the sky
temperature, Ts, rather than the ambient temperature. For the estimation of Ts,
data plot (Fig. 5 in Ref. [37]), in which the scatter of experimental
the simple relation Ts ¼ 0.0552 T 1:5
a was used [40]. points spans the region bounded by DT ¼ 6–18 K at GT ¼ 600 W/m2
7
For vertical surfaces the numerical constant is 1.31 [42,43] and DT ¼ 10–27 K at GT ¼ 1000 W/m2. The experimental average
 ARTICLE IN PRESS

1398 E. Skoplaki et al. / Solar Energy Materials & Solar Cells 92 (2008) 1393–1402

50.0 60.0

45.0
hwind

Heat transfer coefficients hj (W/m2K)

55.0
40.0 hfree

Operating temperature, Tc (°C)

hradn
35.0 Tc 50.0
30.0

25.0 45.0

20.0
40.0
15.0

10.0
35.0
5.0

0.0 30.0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Free-stream wind speed (m/s)

Fig. 1. Relative importance of the individual heat transfer coefficients in the calculation of the PV operating temperature (Ta ¼ 30 1C and GT ¼ 1000 W/m2).

4.0 70.0
Ta=30 degC, Gt=1000 W/m2_with radn
3.5 Ta=30 degC, Gt=1000 W/m2_with only
Operating temperature (°C)

60.0
Ta=10 degC, Gt=600 W/m2_with radn
2
3.0 Ta=10 degC, Gt=600 W/m _with only
50.0
2.5
40.0
hw/hrs

2.0
30.0
1.5
Ta=10degC, Gt=600 W/m2 20.0
1.0 Ta=20degC, Gt=600 W/m2
Ta=30degC, Gt=600 W/m2 10.0
Ta=10degC, Gt=1000 W/m2
0.5 Ta=20degC, Gt=1000 W/m2
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Ta=30degC, Gt=1000 W/m2 Free-stream wind speed, Vf (m/s)
0.0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Fig. 3. Effect of the radiation and free-convection terms on the prediction of PV
Free-stream wind speed (m/s) operating temperature (VwX1 m/s).

Fig. 2. Ratio of the wind convection coefficient to the radiation coefficient for the
extreme combinations of environmental parameters, over the ambient tempera-
ture and irradiance ranges studied.
35

30
Table 2
Relative size of radiation and wind convection coefficients, and the overall effect Exper Avg [37]
25 Vf =1 m/s
on the PV operating temperature prediction of ignoring radiation and free
ΔT=Tc-Ta (°C)

convection Vf =2 m/s
20
Vf =3 m/s
Conditions hw/hrs extremes DT a Vf =4 m/s
15
Vf =5 m/s
Ta ¼ 10 1C (TsE10 1C) 1.01 1.3 10
GT ¼ 600 W/m2 (at Vf ¼ 1 m/s)
Ta ¼ 30 1C (TsE10 1C) 3.34 3.6 5
GT ¼ 1000 W/m2 (at Vf ¼ 15 m/s) (Vf ¼ 1–15 m/s)
(worst case) 0
0 200 400 600 800 1000
[DT ¼ Tc (rad’n, free & forced conv)–Tc (wind conv. only)].
a
Over the wind speed range 1–15 m/s. Solar irradiation (W/m2)

Fig. 4. Predicted operating temperature increase over the ambient one, as a

function of solar irradiance and free-stream wind speed, for free-standing PV
arrays (Eq. (21)).
line is very close to the prediction of Eq. (21) for free-stream wind
speed VfE3 m/s, which corresponds to a surface wind speed
Vw ¼ 2 m/s (According to the Chartered Institute of Building also be noted that early thermal testing of PV modules used to give
Services-CIBS, VwE0.67Vf [47]). This, in a sense, corroborates the higher-slope curves, an average trend-line of which would include
statement that such wind speeds represent ‘‘normal exposure’’ for the points DT ¼ 20 K at GT ¼ 600 W/m2 and DT ¼ 46 K at GT ¼ 1000
an external surface, as pointed out in the previous section. It should W/m2 [36], i.e. it would be above the depicted curves of Fig. 4.
 ARTICLE IN PRESS

E. Skoplaki et al. / Solar Energy Materials & Solar Cells 92 (2008) 1393–1402 1399

0.08 0.070

0.07 0.060
ω=2.4
0.06
(Tc-Ta)/GT (m2K/w)

0.050

(Tc-Ta)/GT (m2K/W)
0.05 ω=1.0 (free-standing)
ω=1.2 (flat-roof) 0.040 Vf =(Vw+0.5)/0.68
0.04 ω=1.8 (slopped-roof) Vf =Vw /0.67
ω=2.4 (façade-integrated) 0.030
0.03

0.02 0.020

0.01 0.010

0.00 0.000
0 2 4 6 8 10 12 14 16 0 5 10 15 20
Free-stream wind speed (m/s) Local wind speed, Vw (m/s)
Fig. 5. Predicted operating temperature increase over the ambient divided by the
Fig. 6. Predicted DT/GT ratio as a function of local wind speed, for fac-ade arrays of
irradiance, as a function of free-stream wind speed and the mounting coefficient
average integration level (o ¼ 2.4).
(Eq. (25)).

Table 3 40
Values of the mounting coefficient, o, for various PV array mounting situations (ω=1.2)
35
PV array mounting type o Exper Avg [37]
30 Vf =1 m/s
Vf =2 m/s
ΔT=Tc-Ta (°C)
Free standing 1.0 25
Flat roof 1.2 Vf =3 m/s
Sloped roof 1.8 (1.0–2.7) 20 Vf =4 m/s
Vf =5 m/s
Fac- ade integrated 2.4 (2.2–2.6)
15

10

If instead of Eq. (20) the analogous relation of Nusselt–Jürges 5

[48] is used for the wind convection coefficient, namely
0
hw ¼ 5:7 þ 3:8V w (22) 0 200 400 600 800 1000
Solar irradiance (W/m2)
Then Eq. (21) is written as
  Fig. 7. Predicted operating temperature increase over the ambient one, as a
0:25
Tc ¼ Ta þ GT (23) function of solar irradiance and free-stream wind speed, for flat-roof mounted PV
5:7 þ 3:8V w arrays (o ¼ 1.2).
where Vw is the local wind speed component parallel to the PV
module’s surface.
As expected, the predictions of Eq. (23) are quite similar, albeit as a (Tc–Ta)/GT versus Vf plot for the various o values of Table 3.
slightly lower, to those of Eq. (21). The fac-ade-integrated prediction of Eq. (25) is further examined in
The use of Eq. (21) as such is limited to free-standing arrays. Fig. 6, which gives results based on local rather than free-stream
However, on the basis of the aforementioned Ross parameter wind speed values. The transformation from Vf to Vw is effected
categorization (cf. Section 5), which was meant to extend the using the relations [45]
validity of Eq. (14) to PV installations beyond the free-standing
V f ¼ ðV w þ 0:5Þ=0:68 (26)
variety, a ‘‘mounting coefficient’’, o, can be defined as the ratio of
the Ross parameter for the mounting situation at hand to the Ross and [47]
parameter for the free-standing case, i.e.
V f ¼ V w =0:67 (27)
kintegration category
o¼ (24) in Eq. (25). It should be noted that selecting the proper wind
kfree-standing array
speed for the estimation of the external forced convection
Using Table 1 values of k and Eq. (24), o values (and ranges) coefficient is a major problem in the research community so that
applying to four types of PV array mounting, namely, free- extra care must be exercised [15]. There are at least seven
standing, flat-roof, sloped-roof, and facade integrated, are listed in definitions in use (‘‘local at surface’’, ‘‘near surface’’, ‘‘roof-top’’,
Table 3. ‘‘above roof’’ on a mast-mounted anemometer, ‘‘30 cm from wall’’,
This normalized (with respect to the free-standing case) ‘‘10 m above ground’’ as in weather stations, and ‘‘near plate’’ in
mounting coefficient can now be used to modify Eq. (21) in order wind tunnel measurements—cf. Refs. [45–47,49]). In the present
to render it valid for situations other than the original free- work, Vw is loosely used to denote local or near the PV array wind
standing case, for which it was developed. Thus, the modified PV speed, while Vf is the speed measured by a mast-mounted
module/array operating temperature is given by anemometer well above the PV array.
  For the purpose of comparison with the trend-lines of the IEA
0:32
Tc ¼ Ta þ o GT (25) PVPS experimental data [37], Figs. 7–9 present DT ¼ Tc– Ta values,
8:91 þ 2:0V f
as a function of irradiance and wind speed, for the three mounting
for any type of mounting and for wind speeds 40 m/s. The types other than the free-standing (or rack-mount, as it is also
resolution offered by the mounting coefficient is shown in Fig. 5, called), and the dashed curves represent the trend-lines in the
 ARTICLE IN PRESS

1400 E. Skoplaki et al. / Solar Energy Materials & Solar Cells 92 (2008) 1393–1402

60 65
Servant [50]
(ω=1.8) Eq.(25) ω=1.0 Ta=20°C, Vf =1 m/s
60

PV operating temperature (°C)

50 Risser & Fuents [53]
Exper Avg [37] King [54]
55
Vf =1 m/s King et al [55]
40 Vf =2 m/s King et al [15]
ΔT=Tc-Ta (°C)

50
Vf =3 m/s Nusselt-Jurges [48]
30 Vf =4 m/s 45
Vf =5 m/s
40
20
35
10 30

0 25
0 200 400 600 800 1000 300 400 500 600 700 800 900 1000

Solar irradiance (W/m2) Solar irradiance (W/m2)

Fig. 8. Predicted operating temperature increase over the ambient one, as a Fig. 10. Operating temperature as a function of irradiance predictions of the
function of solar irradiance and free-stream wind speed, for typical sloped-roof proposed equation compared to explicit correlations which also include the
mounted PV arrays (o ¼ 1.8). ambient temperature and wind speed as variables (Ta ¼ 20 1C, Vf ¼ 1 m/s). For
reasons of clarity, two more correlations [51,52] are not depicted, as they almost
fall on the King et al. [16] curve. Lowest predictions in Servant [50], highest in
80 Nusselt and Jürges [48] and Risser and Fuentes [53].
(ω=2.4)
70
Exper Avg [37] 55
60 Vf =1 m/s
Vf =2 m/s 50
ΔT=Tc-Ta (°C)

PV operating temperature (°C)

50 Ta=20 deg. C
Vf =3 m/s
Vf =4 m/s
40 45
Vf =5 m/s
30 40
20
35
10

0 30
0 200 400 600 800 1000
20
Solar irradiance (W/m2)

Fig. 9. Predicted operating temperature increase over the ambient one, as a 25

function of solar irradiance and free-stream wind speed, for typical fac- ade- 300 400 500 600 700 800 900 1000
mounted PV arrays (o ¼ 2.4). Solar irradiance (W/m2)

Fig. 11. Operating temperature versus solar irradiance as predicted by the latest
correlation of King et al. [16] and Eq. (25) with o ¼ 1.0, for wind speeds of 1, 3, and
relevant experimental data plots. In Fig. 7, which refers to flat-roof 5 m/s (Ta ¼ 20 1C).
PV mounts, the experimental trend-line is that of Fig. 5 in
Ref. [37], where the scatter of experimental points8 is DT ¼ 7–20 K
at GT ¼ 600 W/m2 and DT ¼ 11–28 K at GT ¼ 1000 W/m2. Thus, the 0.0350
curves of Fig. 7 are practically all inside the region bounded by the Eq.(25) ω=1
0.0300 Eq.(23) - Nusselt-Jurges hw
experimental points’ scatter. The same is true for the central area
King [54]
of the plots in Fig. 8 (sloped-roof PV mounts, experimental scatter King et al. [55]
0.0250
from Fig. 4 of Ref. [37]: DT ¼ 10–30 K at GT ¼ 600 W/m2) and Fig. 9 King et al. [16]
(fac- ade-integrated PV’s, experimental scatter from Fig. 3 of
(Tc-Ta)/GT

0.0200
Ref. [37]: DT ¼ 20–40 K at GT ¼ 600 W/m2), for which the experi-
mental data are abundant. However, such good agreement 0.0150
between the predictions of Eq. (25) with the particular experi-
mental data is obviously to be expected, as the development of the 0.0100
correlation has been based on the average behavior of that
0.0050
particular data.
The proposed equation is compared in Fig. 10 to other explicit
0.0000
correlations which also give the PV operating temperature as a 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
function of irradiance, ambient temperature, and wind speed. In Free-stream wind speed (m/s)
this Tc vs GT plot for a single set of Ta ¼ 20 1C, and Vf ¼ 1 m/s
values, the predictions of Eq. (25) for o ¼ 1.0 are relatively close to Fig. 12. Comparison of the proposed equation predictions of the DT/GT ratio
the King [54] and King et al. curves [16,55]. Therefore, in Fig. 11 a dependence on the free-stream wind speed, with other analogous correlations.

more detailed comparison is given between the latest correlation

of King et al. [16] and Eq. (25) with o ¼ 1.0, for three wind speeds,
namely, 1, 3, and 5 m/s, and for the same ambient temperature.
8
The scatter of experimental points is attributed to ‘‘influences other than the The corresponding curves differ by 2–18.2% over the irradiance
irradiation, such as wind or airflow’’ [37] and wind speed ranges shown, at Ta ¼ 20 1C. For the same
 ARTICLE IN PRESS

E. Skoplaki et al. / Solar Energy Materials & Solar Cells 92 (2008) 1393–1402 1401

Table 4
Comparison of Eq. (25) predictions with selected experimental data

GT (W/m2) Vf (m/s) o DT ¼ Tc–Ta (1C) Exper. (1C) |DT diff| (1C) % diff Exper. data source

600 1.2 1.0 17.0 17.0 0 0 [58]

853 1.2 1.0 24.1 28.0 3.9 13.9 [59]
853 2.0 1.0 21.1 23.0 1.9 8.3 [59]
853 3.0 1.0 18.3 20.5 2.2 10.7 [59]
853 4.0 1.0 16.1 15.5 0.6 4.0 [59]
920 1.2 1.0 26.0 25.0 1.0 4.0 [58]
1100 2.0 1.0 27.3 26.0 1.3 4.6 [6]
Average 1.6 6.5

0.120 0.122
0.118
0.116 0.120
Electrical efficiency, ηc

Electrical efficiency, ηc
0.114
0.118
0.112
0.110
0.116
0.108
0.106 Eq. (28) 0.114
Siegel et al. [56] Eq. (28)
0.104
Kou et al. [57] adapt. from ref. [25]
0.112
0.102 adapt. from ref. [25]
0.100
300 400 500 600 700 800 900 1000 0.110
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Solar irradiance (w/m2)
Free-stream wind velocity (m/s)
Fig. 13. Comparison of predicted electrical efficiency as a function of solar
Fig. 14. Comparison of predicted electrical efficiency as a function of wind speed
irradiance with other analogous correlations (Ta ¼ 20 1C, Vf ¼ 1 m/s). That of Kou et
with a correlation adapted from Ref. [25] (Ta ¼ 20 1C, GT ¼ 600 W/m2).
al. [57] is closest.

irradiance and wind speed ranges but for ambient temperatures of where A is the (aperture) surface area of the module. Fig. 13
10 1C and 30 1C, the deviation ranges are 3.1–23.4% and 1.5–14.9%, compares the predictions of Eq. (28) with similar expressions in
respectively. the literature, including a correlation which results from a
The wind speed contribution term in the King/King et al. combination of Eq. (5) of this work with an expression for Tc in
correlations [16,54,55] can be isolated, just as in Eq. (25). Thus, a Ref. [25], i.e. Eq. (23). 3.4. It can be seen that the maximum
further comparison in the form of a Tc– Ta/GT versus Vf plot is deviation of Eq. (28) predictions from the lowest curve in the
shown in Fig. 12, in which Eq. (23) predictions have also been figure is less than 6.5%, at an irradiance level of 1000 W/m2.
included, exhibiting a faster decay of the ratio, especially in the In view of the fact that the Siegel et al. correlation [56] is known to
slow speeds’ region. under-predict Zc by as much as 5%, it can be safely assumed that
The operating temperature rise over the ambient, as predicted the predictions of Eq. (28) are quite satisfactory. The same general
by the proposed correlation (25), is further compared in Table 4 to behavior can be observed in a Zc versus Ta plot, although the drop
experimental values from various sources for the free-standing in Zc is somewhat faster. Finally, Fig. 14 depicts the expected
case, for which the pertinent quantities are available. It can be favorable influence of the wind upon Zc. Thus, the electrical
seen that the absolute differences between the predictions and efficiency predicted by Eq. (28) is a slowly decreasing function of
the measurements average to 1.6 1C, while the corresponding GT, it drops rather quickly with Ta, but increases steeply with Vf in
average percent error is 6.5%. It should be noted that a 5% error in the low wind speed region, leveling off at higher speeds.
Tc results in a 3% error in power output prediction [54]. Moreover,
complicated simulators like PVFORM [28] predict Tc within 5 1C
from the corresponding experimental value. 8. Conclusions
Finally, substitution of the simplified expression of Eq. (25) for
Tc back into Eq. (5) leads to an analogous simple empirical The operating temperature plays a central role in the photo-
expression for the electrical efficiency of a typical crystalline voltaic conversion process, as both the electrical efficiency
silicon PV module, in the form and—hence—the power output of a PV module depend practically
   linearly but rather strongly on Tc, decreasing with it. The pro-
0:32
Zc ¼ 0:12 1  0:004 T a þ o GT  25 (28) posed simple working equations for the PV operating temperature—
8:91 þ 2:0V f Eq. (25)—and for the electrical efficiency/power—Eqs. (28,29)—
with Ta in 1C. The equivalent correlation for the electrical power apply to PV arrays of arbitrary mounting. They explicitly involve
output is then three basic and readily available environmental variables, namely,
   solar irradiance, ambient temperature, and wind speed, as well as a
0:32 dimensionless mounting parameter that is a measure of the
P m ¼ 0:12AGT 1  0:004 T a þ o GT  25 (29)
8:91 þ 2:0V f integration level of a given installation. Use of this parameter
 ARTICLE IN PRESS

1402 E. Skoplaki et al. / Solar Energy Materials & Solar Cells 92 (2008) 1393–1402

makes possible the extension to other integration schemes of an [25] J.A. Duffie, W.A. Beckman, Solar Energy Thermal Processes, third ed., Wiley,
operating temperature correlation developed for free-standing Hoboken, NJ, 2006.
[26] M.C. Alonso Garcia, J.L. Balenzategui, Renew. Energy 29 (2004) 1997.
arrays. The proposed correlations ignore free-convection, which is [27] M.K. Fuentes, A simplified thermal model for flat-plate photovoltaic arrays,
small for all wind speeds, and radiation, which is important only at Report SAND-85-0330, 1987.
wind speeds o1 m/s and, therefore, should be used for wind [28] D.F. Menicucci, J.P. Fernandez, User’s Manual for PVFORM, Report # SAND85-
0376-UC-276, Sandia Natl. Labs, Albuquerque, NM, 1988.
speedsX1 m/s. The average deviation of predicted Tc values from [29] M.Z. Hossain, A.K.M. Sadrul Islam, D.G. Infield, ISESCO Sci. Technol. Vision 3
experiment is less than 3 1C over the wind speed range studied, i.e. (4) (2007) 59.
1–15 m/s. In the case of free-standing arrays, this difference can go [30] B. Marion, B. Kroposki, K. Emery, J. del Cueto, D. Myers, C. Osterwald,
Validation of a photovoltaic module energy ratings procedure, National
down to about 1.6 1C or 6.5% in the 1.2–4 m/s wind speed range, for Renewable Energy Laboratory Report, NREL/TP-520-26909, 1999.
irradiance values between 600 and 1100 W/m2. [31] R.P. Kenny, G. Friesen, D. Chianese, A. Bernasconi, E.D. Dunlop, Energy rating
of PV modules: comparison of methods and approach, presentation at the
Third World Conference on Photovoltaic Energy Conversion, May 11–18,
References Osaka, Japan, 2003.
[32] M. Sandberg, B. Moshfegh, Build. Environ. 37 (2002) 211.
[1] J.S. Griffith, N.S. Rathod, J. Paslaski, Some tests of flat plate photovoltaic [33] B.J. Brinkworth, R.H. Marshall, Z. Ibarahim, Solar Energy 69 (2000) 67.
module cell temperatures in simulated field conditions, in: Proceedings of the [34] L. Mei, D. Infield, U. Eicker, V. Fux, Energy Build. 35 (2003) 605.
15th IEEE Photovoltaic Specialists Conference, May 12–15, Kissimmee, FL, [35] R.G. Ross, Interface design considerations for terrestrial solar cell modules, in:
822, 1981. Proceedings of the 12th IEEE Photovoltaic Specialists Conference, Nov. 15–18,
[2] M. Buresch, Photovoltaic Energy Systems, McGraw-Hill, 1983. Baton Rouge, LA, USA, 1976, pp. 801–806.
[3] H.A. Zondag, Flat-plate PV-Thermal collectors and systems A review, Renew. [36] R.G. Ross, M.I. Smokler, Flat-plate solar array project final report—Vol. VI:
Sustain Energy Rev., 2007. doi:10.1016/j.rser.2005.12.012. Engineering Sciences and reliability, in: Report DOE/JPL-1012-125, 1986.
[4] ASTM, E 1036-02, 2002, Standard test methods for electrical performance of [37] T. Nordmann, L. Clavadetscher, Understanding temperature effects on
non-concentrator terrestrial photovoltaic modules and arrays using reference PV system performance, in: Proceedings of the Third World Conference
cells—Annex A.1 Method for determining the nominal operating cell on Photovoltaic Energy Conversion, May 11–18, Osaka, Japan, 2003,
temperature (NOCT) of an array or module, p. 6. pp. 2243–2246.
[5] W. De Soto, Improvement and validation of a model for photovoltaic array [38] T. Markvart (Ed.), Solar Electricity, Second ed., Wiley, Chichester, 2000.
performance, M.S. Thesis, Mechanical Engineering, University Wisconsin- [39] P.I. Cooper, Solar Energy 27 (5) (1981) 413.
Madison, 2004. [40] W.C. Swinbank, Q. J. Roy. Meteorol. Soc. 89 (1963) 339.
[6] D.L. King, J.A. Kratochvil, W.E. Boyson, Temperature coefficients for PV [41] D.Y. Goswami, F. Kreith, J.F. Kreider, Principles of Solar Engineering, Second
modules and arrays: measurement methods, difficulties, and results, in: ed., Taylor Francis, Philadelphia, 2000, p. 98.
Proceedings of the 26th IEEE Photovoltaic Specialists Conference, September [42] ASHRAE, Handbook of Fundamentals, Atlanta, 1997, p. 3.12.
29–October. 3, Anaheim CA, USA. 1997. [43] A.D. Jones, C.P. Underwood, Solar Energy 70 (2001) 349.
[7] B. Kroposki, D. Myers, K. Emery, C. Whitaker, J. Newmiller, presentation at the [44] U. Eicker, Solar Technologies for Buildings, Wiley, Chichester, 2003.
25th IEEE Photovoltaic Specialists Conference, May 13–17, Washington DC, [45] D.L. Loveday, A.H. Taki, Int. J. Heat Mass Trans. 39 (8) (1996) 1729.
USA, 1996. [46] M.G. Davies, Building Heat Transfer, Wiley, Chichester, 2004, p. 113.
[8] D.L. Evans, Solar Energy 27 (1981) 555. [47] S. Sharples, Build. Environ. 19 (1984) 31.
[9] G. Notton, C. Cristofari, M. Mattei, P. Poggi, Appl. Therm. Eng. 25 (2005) 2854. [48] W. Nusselt, W. Jürges, Die Kühlung einer ebenen Wand durch einen Luftstrom
[10] D.L. Evans, L.W. Florschuetz, Solar Energy 19 (1977) 255. The cooling of a plane wall by an air flow, Gesundheits Ingenieur 52, Heft 45,
[11] G.W. Hart, P. Raghuraman, Simulation of thermal aspects of residential Jahrgang 30 Dezember 1922, pp. 641–642.
photovoltaic systems, MIT Report DOE/ET/20279-202, 1982. [49] N. Ito, K. Kimura, J. Oka, ASHRAE Trans. 78 (1972) 184.
[12] H.P. Garg, R.K. Agarwal, Energy Conv. Manage. 36 (2) (1995) 87. [50] J.M. Servant, Proceedings of the Ninth Bienn. Congr. ISES, Pergamon, Oxford,
[13] D.L. Evans, L.W. Florschuetz, Solar Energy 20 (1978) 37. 1986.
[14] G. Sala, Cooling of solar cells, in: A. Luque, G.L. Araujo (Eds.), Solar Cells [51] M. Mattei, G. Notton, C. Cristofari, M. Muselli, P. Poggi, Renew. Energy 31
and Optics for Photovoltaic Concentration, Adam Hilger–IOP Publishing, 1989, (2006) 553.
pp. 239–267. [52] R. Chenni, M. Makhlouf, T. Kerbache, A. Bouzid, Energy 32 (2007) 1724.
[15] J.A. Palyvos, Appl. Therm. Eng. 28 (2008) 801. [53] V.V. Risser, M.K. Fuentes, Linear regression analysis of flat-plate photovoltaic
[16] D.L. King, W.E. Boyson, J.A. Kratochvil, Photovoltaic array performance model, system performance data, in: Proceedings of the Fifth E.C. Photovoltaic Solar
/http://www.sandia.gov/pv/docs/PDF/King%20SANDS. pdf, Rept. SAND2004- Energy Conference, Oct. 17–21, 1983, Athens, Greece, pp. 623–627.
3535, 2004. [54] D.L. King, Photovoltaic module and array performance characterization
[17] J.W. Stultz, L.C. Wen, Thermal performance testing and analysis of photo- methods for all system operating conditions, in: Proceedings of NREL/SNL
voltaic modules in natural sunlight, DOE/JPL LSA Task Report 5101-31, 1977. Photovoltaic Program Review Meeting, Nov. 18–22, Lakewood, CO, USA, 1997.
[18] A. Kirpich, G. O’Brien, N. Shepard, Electric power generation: photovoltaics, pp. 1–22.
in: W.C. Dickinson, P.N. Cheremisinoff (Eds.), Solar Energy Technology [55] D.L. King, J.A. Kratochvil, W.E. Boyson, W.I. Bower, Field Experience with a
Handbook Part B, Marcel Dekker, New York, 1980, p. 329. new performance characterization procedure for photovoltaic arrays, in:
[19] Florida Solar Energy Center, Test method for photovoltaic module power Proceedings of the Second World Conference on Exhibition on Photovoltaic
rating, FSEC Standard 202-05, May 2005, /www.fsec.ucf.eduS. Solar Energy Conversion, Vienna, 1998, pp. 1947–1952.
[20] J.J. Bloem, Build. Environ. 43 (2008) 205. [56] M.D. Siegel, S.A. Klein, W.A. Beckman, Solar Energy 26 (1981) 413.
[21] J.W. Stultz, Thermal and other tests of photovoltaic modules performed in [57] Q. Kou, S.A. Klein, W.A. Beckman, Solar Energy 64 (1–3) (1998) 33.
natural sunlight, In: Report DOE/JPL-1012-78/9, 1978. [58] J.J. Bloem, W. Zaaiman, An autonomous monitoring device for PV Installations
[22] J.H. Eckstein, Detailed modelling of photovoltaic system components, and PV calculation validation, In: Proceedings of the 19th European
M.S. Thesis, Mechanical Engineering, University of Wisconsin-Madison, 1990. Photovoltaic Solar Energy Conference and Exhibition, June 7–11, 2004, Paris,
[23] T.U. Townsend, A Method for estimating the long-term performance of direct- France.
coupled photovoltaic systems, M.S. Thesis, Mechanical Engineering, Uni- [59] K. Furushima, Y. Nawata, M. Sadatomi, Prediction of photovoltaic power
versity of Wisconsin-Madison, 1989. output considering weather effects, In: ASES Conference SOLAR 2006–
[24] M.W. Davis, B.P. Dougherty, A.H. Fanney, Trans. ASME J. Solar Energy Eng. 123 Renewable Energy Key to Climate Recovery, July 7–13, 2006, Denver, CO,
(2) (2001) 200. USA.