Available online at www.sciencedirect.

com

ScienceDirect
Energy Procedia 62 (2014) 694 – 703

6th International Conference on Sustainability in Energy and Buildings, SEB-14

A Comparative Analysis of PV Module Temperature Models

Tuza A. Olukan, Mahieddine Emziane*
Solar Energy Materials and Devices Laboratory
Masdar Institute of Science and Technology, PO Box 54224, Masdar City, Abu Dhabi,UAE

Abstract

The purpose of this work is to show the impact of various thermal models used in predicting the module temperature on the
power output of PV modules installed in the UAE. As it is known, the module temperature of PV affects its electrical power
output. In this regard we developed a three-dimensional thermal model to predict the temperature of PV modules using finite
element method (FEM). Our model was compared with some simple thermal models used by certain software packages and
simple correlations in the literatures to predict the module temperature of PV. We used field data from our PV testing facility to
validate both the thermal and electrical output of all models used in this work. From the results, all the thermal models were
found to underestimate the module temperature of PV operating in the UAE. When the thermal models were used in the two-
diode model with other parameters to predict the power output of the module, the results show that they all overestimate the
power output with our model being the most accurate and the least biased among all the models considered.

© 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
© 2014 The Authors. Published by Elsevier Ltd.
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Selection and peer-review under responsibility of KES International.
Selection and peer-review under responsibility of KES International

Keywords: Matlab; Photovoltaic; Two-diode model; temperature; power output

1. INTRODUCTION

Considering the peculiar nature of the UAE environment, its obvious solar energy potential, and in line with the
government policy of achieving a certain proportion of its energy from renewable source, harnessing the abundantly

* Corresponding author. Tel.: +971-02-810-9140;

E-mail address: memziane@masdar.ac.ae

1876-6102 © 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Selection and peer-review under responsibility of KES International
doi:10.1016/j.egypro.2014.12.433
 Tuza A. Olukan and Mahieddine Emziane / Energy Procedia 62 (2014) 694 – 703 695

available solar resources is undoubtedly a viable mean of achieving these lofty goals. Hence a comprehensive
research work in this area is inevitable and surely a right trajectory to this target.

In line with the foregoing, the ability to replicate real-world processes and procedures in a virtual environment can
help prone down the cost of studying and maintaining such a system in reality. In addition it also gives the engineer
flexibility in carrying out research on such a system. With this mindset, this project intends replicating a typical solar
PV module installed in UAE with the aid of Matlab tool. In this work, we use the two-diode model to predict the
power output of an installed PV module in the UAE. This model was chosen because it is more accurate than the
single diode model especially at low solar irradiance [1, 2].

Critical to the output of this model are the module temperature and the solar irradiance falling on the surface of the
PV module. This work examines various models used in predicting the module temperature of PV. These models are
the Ross temperature model, the temperature model used by Homers’ PV software, the Faiman PV module
temperature model used by the commercial modeling PV software package, PVSYST [3-7] and a three dimensional
temperature model proposed in our recent work [8]. It was shown that our proposed model is the most accurate and
was therefore used to predict the module temperature in the two-diode model and subsequently the power output of a
PV module operating in UAE environment conditions.

2. METHODOLOGY

We will examine all the temperature models mentioned earlier and validate them against real data from outdoor PV
testing facility that we have in our Institute. These data include the performance (power output and module
temperature) of poly-crystalline Si module and metrological data of the location as well, all captured at a resolution of
10 minutes. The metrological data, i.e. the solar irradiance and the module temperature are used as input parameters
into the two-diode model implemented in Matlab.

2.1 TEMPERATURE MODELS

2.1.1 Ross thermal Model

The most common expression for finding the cell temperature of PV module is by using the normal operating
condition temperature (NOCT) of the PV module with the relation by Ross expressed as:

ሺTNOCT െʹͲሻ
C ൌ a ൅ ͺͲͲ
ൈ
T (1)

This equation is valid only for free standing PV modules [5]. The value of ambient temperature and solar irradiance
on the plane of array are provided from our field data and NOCT of 44 qC was used.
696 Tuza A. Olukan and Mahieddine Emziane / Energy Procedia 62 (2014) 694 – 703

2.1.2 Temperature Model used by Homer Software

According to the information on Homer help page, the software evaluates the PV cell temperature by the relation in
equation (2) [6]. The following values, 44°C, 20°C, and 25 °C were used for T C,NOCT , Ta,NOCT , and TC,STC
respectively. GT, NOCT value is 800W/m2. The value of 0.9 was used for τα in the equation while the electrical
conversion efficency (12%) of the PV modules was used. The wind speed was not taken into account in this
equation. The temperature coefficient used in the computation was extracted from the manufacturer’s data sheet.

(2)

2.1.3 PVSYST Temperature Model

PVSYST evaluates PV cell temperature according to the Faiman module temperature model [9]. This model is
expressed by the equation below:

(3)

Usually, a default value of 0.9 is fixed for the absorption coefficient in PVSYST, this default value was used in
computing the cell temperature in this work. Since the efficiency of the module considered in this work is known it
was subsituted for the PVSYST default value of 0.1. For the constant heat transfer component in the expression, the
PVSYST default value of 29 W/m2 was used for free standing mounting. The software doesn’t ellaborate on how the
convective heat transfer component is handled in the model, hence the default value (U 1=0) was used in the
computation of the module temperature [9]. Unlike other models considered, this model does not require the NOCT
temperature of the module for the module temperature computation and the software advised against using this value
for computing the PV module temperature [7].

2.1.4 Proposed Temperature Model

We model the surface temperature of a PV module using finite element method (FEM) through Comsol Multiphysics.
More details are reported in our recent work [8]. The proposed temperature model is different from the earlier three
models described above because it predicts the surface temperature of PV modules during periods of rapidly
changing irradiance where the response time due to the thermal mass of the PV material comes in to play significantly
[10]. In addition, these are more suitable for the rack-mounted installations and make the approximations that the
overall heat loss coefficient is constant [11]. Our proposed model calculates the non-steady state thermal response of
 Tuza A. Olukan and Mahieddine Emziane / Energy Procedia 62 (2014) 694 – 703 697

the PV module and can also take into consideration different mounting conditions from the one considered in the
other simple thermal models. Our model also uses different heat loss coefficients depending on the wind speed and
direction. These values are necessary to calculate the module surface temperature, as both sides of the PV module
might be subjected to different conditions. Furthermore, this method ensures that the dynamic thermal response of the
PV module under the fluctuating weather conditions is well captured. This method of calculating the surface
temperature is imperative, as the temperature of a PV module is sensitive to its electrical output characteristics and
efficiency. Hence, this ensures that the input surface temperature into the two-diode model represents the actual
operating conditions of PV in the UAE environmental conditions.

2.2 MODEL OF A PV CELL

Figure 1: Two diode model of a PV cell [2].

A PV cell can be modelled as shown Figure 1. This model takes into account the recombination loss in the depletion
region which is neglected in the single diode model [12]. This inclusion however comes at price as more parameters
will have to be computed and this increases the computation time as compared to the single diode model [1]. From the
circuit diagram in Figure 1, the relationship between the current output and terminal voltage can be expressed as:

ሺ୚ାRS ൈ୍ሻ
I ൌ Ipv െ ID1 െ ID2 െ RP
(4)

The appearance of I on both sides of the equation means equation (4) is a non-linear equation and has to be solved
using numerical analysis. The Newton-Raphson numerical analysis is selected to solve the equation. The light-
generated current depends linearly on the solar irradiation and temperature according to equation below (5).

ீ
‫ ݄݌ܫ‬ൌ ሺ‫݄݌ܫ‬ǡ ݊ ൅ ‫ݐ݀݅ܭ‬ሻ (5)
ீ௡
698 Tuza A. Olukan and Mahieddine Emziane / Energy Procedia 62 (2014) 694 – 703

ሺோ௦ାோ௣ሻ
‫݄݌ܫ‬ǡ ݊ ൌ ‫ܿݏܫ כ‬ǡ ݊ (6)
ோ௣

‫ܿݏܫ‬ǡ ݊ is the short-circuit current at standard testing condition (STC) and is provided by the manufacturer. With every
parameter defined, next is to define other parameters in equation (4) i.e. ‫ʹܦܫ݀݊ܽͳܦܫ‬:

௏ାோ௦‫כ‬ூ
‫ ͳܦܫ‬ൌ ‫ͳͲܫ‬ሾ݁‫݌ݔ‬ሺ ሻ െ ͳሿ (7)
௏௧‫כ‬௔ଵ

௏ାோ௦‫כ‬ூ
‫ ͳܦܫ‬ൌ ‫ʹͲܫ‬ሾ݁‫݌ݔ‬ሺ ሻ െ ͳሿ (8)
௏௧‫כ‬௔ଶ

‫ ͳͲܫ‬and ‫ ʹͲܫ‬are the saturation currents and are assumed to be equal. a1 and a2 are the ideality factors, with values of
1 and 2 respectively [1]. These values are an approximation of the Shockley-Read-Hall recombination in the space
charge layer of the photodiode [2]. Vt in equations (7) and (8) is defined by:

ሺNS ൈൈሻ
Vt ൌ q
(9)

The saturation current in expressions (7) and (8) can be derived from equation (10):

ሺIphn ൅Ki ݀‫ݐ‬ሻ

I01 ൌ I02 ൌ (10)
൅Kv ݀‫ݐ‬
݁‫݌ݔ‬൦ Vocn
a ൅a ൪െͳ
ቊቆ 1 2 ቇቋVt
p

With all the parameters defined except Rs and Rp, equation (4) can now be re-written as shown in equation (11) :

୚ାRS୍ ୚ାRS୍ ሺ୚ାRS୍ሻ

‫ ܫ‬ൌ Ipv െ  I01 ቂ݁‫ ݌ݔ‬ቀ ቁ െ ͳቃ െ I02 ቂ݁‫ ݌ݔ‬ቀ ቁ െ ͳቃ െ (11)
Vt ൈa1 Vt ൈa2 RP

The values of Rs and Rp, i.e. the series and shunt resistances, have been obtained through iterations by calculating
both values simultaneously [2]. This is done by matching the calculated peak power and the one given by the
manufacturer by increasing RS while computing the value of RP at the same time. This is experimented by using
equation (12) where the initial conditions are given in equation (13). The values of Rs and Rp are reached when the
values of both powers are equal:
 Tuza A. Olukan and Mahieddine Emziane / Energy Procedia 62 (2014) 694 – 703 699

௏௠௣ሺ௏௠௣ାூ௠ோ௦ሻ
ܴ‫ ݌‬ൌ ሾ௏௠௣ሼூ௣௛ିூ஽ଵିூ஽ଶሽି௉௠௔௫ǡ௘ሿ (12)

௏௠௣ ௏௢௖ǡ௡ି௏௠௣
ܴ‫ ݋ݏ‬ൌ ͲǢ ܴ‫ ݋݌‬ൌ ூ௦௖ǡ௡ିூ௠௣ െ ூ௠௣
(13)

With all the parameters defined, the output current can now be calculated using the Newton-Rapson numerical
method.
2.3 TESTING FACILITY

The outdoor performance parameters of the PV module were captured by an IV measurement system, CaptureStar
1100, onTest with appropriate sensors. The information collected consists of the Fill Factor (FF), Maximum Power
(Pmax), Short Circuit current (Isc), Output Voltage (Voc), PV module and ambient temperatures, solar irradiance,
wind direction and wind velocity, humidity, voltage and current of the PV module mounted on a stand on the roof of
our institute building. A suitable location was chosen to avoid shading effects. A pyranometer was connected to the
setup with the following specifications- ISO second class, wavelength range: 305-2800nm, sensitivity: approx.
7mV/kW/m2, and used to collect the outdoor solar irradiance, while a thermocouple was installed underneath the PV
module.

3. RESULTS AND DISCUSSION

It has been widely reported in the literature that an increase in temperature has detrimental effect on the power output
of PV modules. All the thermal models were used in evaluating the cell temperature in the two diode model and the
electrical power output of the module was calculated. Figure 2 shows the result of the thermal models compared to
the actual value of the installed p-Si module. All the models clearly underestimate the actual temperature of the PV
module. The root mean square error and the mean bias error were used as metrics to compare the models accurately
where the two metrics are given by equations (14) and (15). The value of y is the model temperatures; x is the
experimental value while n is the number of data sample. RMSE is used to compare the accuracy of the models with
the actual value, while the MBE is used to gauge how the model over-predicts or under-predicts the models. Figure 3
shows the errors of all the models. The Ross thermal model is the most conservative among the model and the most
biased while our proposed thermal model is the most accurate and the least bias. The most accurate of the simple
thermal models is the Homer temperature model. The results from the Homer thermal model and our proposed model
are used in the two-diode model to predict the electrical power output of the PV module. Figure 4 shows the results of
the power output.

೙ మ
ඨ෍ ൫yi ିxi ൯
೔సభ
ܴ‫ ܧܵܯ‬ൌ σ೙
(14)
೔సభ xi
700 Tuza A. Olukan and Mahieddine Emziane / Energy Procedia 62 (2014) 694 – 703

೙
෍ ൫yi ିxi ൯
‫ ܧܤܯ‬ൌ ೔సభ
೙
σ೔సభ xi
(15)

Figure 2: Temperature predictions by the different models against the experimental results.

Figure 3: Temperature errors for all the used thermal models.

 Tuza A. Olukan and Mahieddine Emziane / Energy Procedia 62 (2014) 694 – 703 701

Figure 4: Maximum power output using Homer model and our proposed model.

From Figure 4, it is found that all the models over -estimate the electrical power production of the PV module when
compared to the actual PV power recorded. As expected, the Homer thermal model over estimates the predicted
electrical power output higher than our proposed model. This is because the model under-estimates the PV module
temperature the most as seen in both Figures 2 and 3. Using our proposed model gives the closest result to the actual
power output recorded in the field. This is expected as our proposed model predicted the temperature of the PV
module with more accuracy compared to the other models. Figure 5 shows the RMSE and MBE of the PV power
output using the various models. All the models over predict the module power output module with quite high errors,
with our model being the most accurate.

Figure 5: Maximum power error for the models considered.

702 Tuza A. Olukan and Mahieddine Emziane / Energy Procedia 62 (2014) 694 – 703

4. CONCLUSION

The thermal model used in predicting the temperature of PV module is important as the module temperature affects
its power output. Four thermal models where used in predicting the module temperatures of PV modules operating in
the UAE environmental conditions. Though all the considered thermal models under-estimate the actual module
temperature, our proposed model predicts the module temperature better and when, used in the two-diode model, also
gives the most accurate electrical power output of the PV modules. It is recommended that existing simple thermal
models need to be improved so as to be able to predict more accurately the temperature and power output of PV
modules.

Nomenclature

I Output current (A)

IPH Generated current (A)
ID1 Shockley diode equation due to diffusion
ID2 Shockley diode equation due to charge recombination mechanisms
V Terminal voltage (V)
Rs Series resistance (Ω)
Rp Shunt resistance (Ω)
Iphn Generated current of the cell at standard testing conditions
Ki Short-circuit current coefficient
dt Change in temperature G Solar irradiation in a particular location in (W/m2)
Gn 1000W/m2 solar irradiation at air mass 1.5 at the standard testing conditions
Ns Number of cells in series
K Boltzmann constant [1.3806503 × 10-23]
q The electron charge [1.602174646 × 10-19 C]
TC Solar radiation striking on the POA of the PV module W/m2
GT Ambient temperature in °C.
Ta Cell temperature in °C.
TC,NOCT Nominal operating cell temperature in °C
Ta,NOCT Ambient temperature at NOCT in °C
Tc,STC Cell temperature at STC in °C.
GT, NOCT Solar radiation at NOCT W/m2
ηmp Electrical conversion efficiency of the PV module in %.
τ Solar transmittance of any cover over the PV module in %.
αp Temperature coefficient of the PV module in %/°C.
α The adsorption coefficient of the module
EPOA The irradiance incident on the POA of the PV module in W/m2
U0 The constant heat transfer component
U1 The convective heat transfer component
Ws wind speed in m/s
 Tuza A. Olukan and Mahieddine Emziane / Energy Procedia 62 (2014) 694 – 703 703

References:

1 Sah, R.-Y., Noyce, R.N., and Shockley, W.: ‘Carrier generation and recombination in pn junctions and pn junction
characteristics’, Proceedings of the IRE, 1957, 45, (9), pp. 1228-1243
2 Ishaque, K., Salam, Z., and Taheri, H.: ‘Accurate MATLAB simulink PV system simulator based on a two-diode model’,
Journal of Power Electronics, 2011, 11, (2), pp. 179-187
3 Koltay, P., Wenk, J., and Bücher, K.: ‘Outdoor Measurements of NOCT and Simulation of its Influence on Energy
Collection’, in Editor (Ed.)^(Eds.): ‘Book Outdoor Measurements of NOCT and Simulation of its Influence on Energy Collection’
(1998, edn.), pp. 6-10
4 Malik, A., and Damit, S.J.B.H.: ‘Outdoor testing of single crystal silicon solar cells’, Renewable Energy, 2003, 28, (9), pp.
1433-1445
5 Jakhrani, A., Othman, A., Rigit, A., and Samo, S.: ‘Comparison of Solar Photovoltaic Module Temperature Models’,
World Applied Sciences Journal, 2011, 14, pp. 1-8
6 http://support.homerenergy.com/index.php?/Knowledgebase/Article/View/489/0/10487---how-homer-calculates-the-pv-
cell-temperature, accessed April 8th 2014
7 http://files.pvsyst.com/help/index.html?noct_definition.htm, accessed April 8th 2014
8 Tuza, A.,Olukan, M. Emziane .: ‘Three Dimensional thermal Modeling of the Annual Temperature of Four Different
Photovoltaic Modules operated in the UAE Environmental Conditions ’, Renewable Energy, under review.
9 http://pvpmc.org/modeling-steps/cell-temperature-2/cell-temperature-3/pvsyst-cell-temperature-model/, accessed April 8th
2014
10 Jones, A., and Underwood, C.: ‘A thermal model for photovoltaic systems’, Solar Energy, 2001, 70, (4), pp. 349-359
11 Davis, M.W., Fanney, A.H., and Dougherty, B.P.: ‘Prediction of building integrated photovoltaic cell temperatures’,
Journal of Solar Energy Engineering, 2001, 123, (3), pp. 200-210
12 Yordanov, G., Midtgård, O., and Saetre, T.: ‘Two-diode model revisited: Parameters extraction from semi-log plots of IV
data’, in Editor (Ed.)^(Eds.): ‘Book Two-diode model revisited: Parameters extraction from semi-log plots of IV data’ (2010,
edn.).