Energy Reports 7 (2021) 7297–7314

Contents lists available at ScienceDirect

Energy Reports
journal homepage: www.elsevier.com/locate/egyr

Research paper

Performance assessment of a 20 MW photovoltaic power plant in a hot

climate using real data and simulation tools
∗
Said Bentouba a , , Mahmoud Bourouis b , Nadjet Zioui c , Arumugam Pirashanthan a ,
Dhayalan Velauthapillai a
a
Western Norway University of Applied Sciences, Faculty of Engineering and Science, Bergen, 5063, Norway
b
Department of Mechanical Engineering, Universitat Rovira i Virgili, Av. Països Catalans No. 26, 43007 Tarragona, Spain
c
Mechanical Engineering Department, University of Quebec at Trois-Rivieres, QC, Canada

article info a b s t r a c t

Article history: The present study aims to evaluate the aptness of two commercial simulators, HOMER Pro and
Received 8 July 2021 RETScreen Expert, as predictors of the performance of a large-scale photovoltaic power plant designed
Received in revised form 11 October 2021 to deliver up to 20 MW in a hot climate, for which 26 months of real operational data are available.
Accepted 13 October 2021
The power plant is located in the province of Adrar in the south of Algeria and classified as one
Available online 6 November 2021
of the hot regions worldwide. Performance parameters were reference yield, performance ratio,
Keywords: capacity factor, temperature loss and statistical indicators. The results showed that photovoltaic power
Large-scale photovoltaic power plant plant performance depends on cell technology, insolation, and environmental conditions, especially
Real-time monitoring temperature. The deviations between the simulation results and real monitoring data were found to
HOMER pro be smaller in the case of HOMER Pro simulation tool. The total annual energy supplied in 2018 by
RETScreen
the power plant was 36364MWh, whereas RETScreen Expert predicted 42339 MWh, or about 14%
Hot climate
more and HOMER Pro predicted 34508 MWh or about 5.1% less. The influence of temperature on the
power plant output was strong, causing a 40% drop during the summer, due to the limitations of
the polycrystalline cell technology. This needs to be considered in the design of future photovoltaic
power plants to be operated in hot climates. HOMER Pro and RETScreen Expert predicted an average
annual final yield of 5.128 h/day, a module efficiency of 15% and an inverter efficiency of 98%. The
t statistics were 3.75 for HOMER Pro and 6.12 for RETScreen Expert. The analysis shows that the 20
MW photovoltaic plant in hot climate experiences high losses compared to an equivalent plant based
on thin-film photovoltaic cells.
© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

1. Introduction these resources, placing its growing domestic demand for elec-
tricity under serious constraints. Meanwhile, the slump in oil
According to International Renewable Energy Agency (IRENA) prices since the third quarter of 2014 has nearly halved Algeria’s
figures, global renewable energy capacity grew to at least 260 foreign currency reserves. The Renewable Energy and Energy
gigawatts (GW) in 2020, an increase of nearly 50% compared Efficiency Development Plan launched in 2011 and updated in
to 2019, despite the slowdown due to the COVID-19 pandemic. 2015 emphasizes deployment of large-scale solar photovoltaic
Solar energy increased by 22% to 127 GW, and wind energy by installations and onshore wind turbines, both made possible by
18% to 111 GW (IRENA, 2021). The increase in solar capacity is decreasing costs as the technologies advance. Biomass, cogener-
due largely to decreasing photovoltaic system costs. New Energy ation and geothermal technologies were to be added until 2020.
The goal of the program is to install 22 000 MW by 2030 (Fig. 1)
Outlook forecasts that the cost of solar electricity will fall by
which includes large grid-connected solar power plants with a
60% over the next 20 years, and converting solar energy will
total capacity of 13 500 MW (Algeria Ministry of Energy, 2021).
become the cheapest way of producing electricity in most of
Photovoltaic system performance will depend on the tech-
the world by the year 2030 (Pothecary, 2016). Algeria depends
nology used and on the climatic parameters of the power plant
heavily on revenue from the exportation of its massive oil and
site. New-generation cells, components of the installation and the
gas reserves, and over 90% of its electricity is generated from
scale of the integration grid are being investigated to optimize
the design and operation of these power plants. In this article,
∗ Corresponding author. we compare the actual monitored performance of a photovoltaic
E-mail address: sben@hvl.no (S. Bentouba). power generator to simulations by HOMER Pro and RETScreen

https://doi.org/10.1016/j.egyr.2021.10.082
2352-4847/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

The analysis of performance ratio, yield energy, reference en-

ergy, capacity utilization factor and energy efficiency of differ-
ent solar photovoltaic were also analyzed in Kumar and Ku-
mar (2017). Furthermore, a review of the effects of dust accu-
mulation and ambient temperature on PV performance in the
Middle East and North Africa was explored in Hammad et al.
(2018), assessment of 2.5 kW photovoltaic power generator per-
formance under conditions of high losses due to environmental
factors in southern Algeria were also surveyed in Necaibia et al.
(2018). In cases studies, an analysis of the performance of grid-
connected desert-type 1.4 kW photovoltaic plants under Muscat
climatic conditions found that the total annual energy yield was
2217.6 kWh whereas the annual average daily reference yield,
array yield and final yield were respectively 6.36 kWh/kWp-day,
4.56 kWh/kWp-day and 4.10 kWh/kWp-day and that energy pro-
duction was 10% lower when the modules were left uncleaned,
but no details were provided on the influence of temperature (Al-
Badi, 2018). In an evaluation of design aspects of a 10 MW grid-
connected photovoltaic power plant in terms of various types of
power losses (temperature, internal network, power electronics,
Fig. 1. The Algerian renewable energy program. grid connected, etc.) and comparison with simulations by PV Sys-
tem and PV-GIS software, the final yield was 1.96–5.07 h/d and
the annual performance ratio was 86.12% (Kumar and Sudhakar,
Expert. We also discuss the consequences of different internal 2015). In a study of a 5 MW photovoltaic power plant designed
component distributions and grid topologies. for 50 Iranian cities using RETScreen software, the highest ca-
An extensive review of the literature reveals a lack of infor- pacity factor (26.1%) was found at Bushier and the lowest at
mation on the internal topology and integration of large-scale Anzali (16.5%) for an average of 22.27% (Besarati et al., 2013).
photovoltaic power plants into the power grid. Simulators are In a PV-GIS-based comparison of monocrystalline silicon, cad-
used to predict the energy production of the proposed system mium telluride (CdTe) and copper indium selenide (CIS) cells
before integrating any large-scale variable renewable energy. We for 1 MW photoelectric solar power plants in 23 Serbian cities,
CdTe was found most suitable, based on total annual produc-
examine the aptness of RETScreen Expert and HOMER Pro using
tion of electricity (Pavlović et al., 2013). In energy performance
the case of a real 20 MW photovoltaic power plant operating in a
prediction and loss estimation for crystalline photovoltaic 200
hot climate. To the best of our knowledge, this is the first research
kWp arrays in Northern India, the estimated degradation rate
article to describe, display and analyze in detail such a large mass
ranged from −0.6/year to −5%/year, light-induced degradation
of data obtained by monitoring power panel configuration, power
was −2.5%/year and the predicted energy losses ranged from
production on hourly, daily, monthly, and yearly basis, insolation, −1757.724 to −14647.7 kWh/year (Kumar et al., 2019). The pho-
temperature, module characteristics, and cable sections in an tovoltaic array efficiency, inverter efficiency and system efficiency
actual grid connection and to compare these to simulations. The of a 40 kWp GIPV system installed in India evaluated for one
data monitoring period was 26 months, from November 2016 to year were found to be 9.36%, 90.9% and 8.51% respectively (Sat-
December 2018. The effectiveness of the system configuration sangi et al., 2018). Based on the International Photovoltaic Project
is examined in terms of total energy generated, performance Model, the best scenario for a 12 kW photovoltaic power plant
ratio, capacity factor and monthly system efficiency. In the second was the satisfaction of power demand by both solar (27%) and
section of this article, the simulators are evaluated. The total grid electricity (73%), with a minimal reduction in GHG emis-
amount of electrical energy generated and supplied to the load sions of 23 t of CO2 per year (Rashwan et al., 2017). In a study
and the various types of power losses are predicted. Finally, the of photovoltaic power plants (1523 kW and multi-MW) built
total energy flow through the whole system is calculated by in the best locations in the Canaries (Spain) and presumably
predicting energy supplied to the load for one year and compared well managed, simulation results deviated from measured spe-
with data from real-time monitoring. cific yields (237 588 monthly energy values for 2005–2017) by
less than 3% (Guerrero-Lemus et al., 2019). The model was also
2. Related studies used to detect suboptimal plant designs and anomalous specific
yields above the clear sky limit. Recommendations to avoid fu-
ture anomalous specific yields were proposed. Analysis based
Studies of large-scale photovoltaic power plants are relatively
on the standard IEC 61724 guidelines of the performance of six
recent. A comparison of actual performance of a 5 MW grid-
large photovoltaic power plants over several years of operation
connected plant in South India with RETScreen predictions was
with different mounting topologies agreed largely with expected
recently performed (Sundaram et al., 2015), techno-economic results by considering plant location in the South-Central Re-
analysis of grid-interactive solar photovoltaic (PV) projects im- gion of Spain, where the total system efficiency ranging from
plemented under the first phase of India’s national solar mis- 10% to 12% (Martín-Martínez et al., 2019). A grid-connected 3
sion (Purohit and Purohit, 2018). Use of novel algorithm relief MW photovoltaic power plant located in Karnataka State (In-
attribute evaluator to evaluate the relative influences of solar dia) generated an average of 1372 kWh per year per kWp of
radiation and back surface module temperature on predicted installed capacity. This performance was found satisfactory based
daily array yield of a 190 kW power plant using a radial ba- on the normalized comparison with installations in other coun-
sis function neural network for 26 Indian cities was presented tries (Padmavathi and Daniel, 2013). Components used in large-
in Yadav et al. (2018). A structured framework was examined scale photovoltaic power plants including panels, converters and
for evaluating photovoltaic plant integration into weak distribu- transformers as well as their distribution and the associated col-
tion systems in grid connection studies in Susanto et al. (2018). lection grid topologies have been reviewed (Cabrera-Tobar et al.,
7298
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Table 1 Table 2
Daily solar radiation recorded in typical cities in the Algerian desert. Design parameters of the Adrar 20 MVp photovoltaic power plant.
Province Solar radiation in kWh/m2 per day Sunshine in hours per day Design parameter Characteristics
Adrar 5.87 12.13 Module type Poly-crystalline silicon
Bechar 5.38 12.12 Photovoltaic module efficiency 15%
Tamnrasset 5.61 12.10 Orientation and tilt 26.5◦ , south
Tindouf 5.86 12.14 Installation type Fixed
Distance between photovoltaic rows 8 m
Inverters 500 kW
Transformers 1.250 kVA, 47–52 Hz, 315 V/31.5 kV
2016b). Technical, environmental and economic aspects of select-
ing 44 sites for 10 MW installed capacity grid-connected photo- Table 3
voltaic power plants in Saudi Arabia have been evaluated using Configuration of the Adrar 20 MWp power plant (SKTM, 2019).
RETScreen to predict energy production, greenhouse gas (GHG) Equipment Number
emissions and financial parameters (Rehman et al., 2017). It was 245 Wp modules 81 840
found that the town of Bisha was best suited due to insola- 1 MW subfields 20
tion (intensity and duration). Design solutions for increasing the Inverters 40
1250 kVA/630 kVA transformers 20
efficiency and yield of photovoltaic power generators mounted
on a floating platform were tested in experiments that showed
a considerable increase in efficiency due to positive tracking
and reduced temperature effects (Cazzaniga et al., 2018). In a 3.2. Description of the 20 MW photovoltaic power plant
study of large-scale photovoltaic systems based on 1006.74 kWp
crystalline silicon (c-Si) arrays in semi-arid climates of India, The photovoltaic power plant studied in this article is situated
the degradation rate calculated from 4 years of operational data in the southwest of Algeria, a region very rich in solar power
(using a linear least square fitting method) ranged from −0.30% potential due to its topography and low latitude. The Adrar sta-
tion is the legal entity mandated to provide renewable energy
to −0.17% starting from the third year of operation, which is
in Algeria, where it has a capacity of 20 MWp and it is one of
within ranges reported previously in similar studies conducted
the seven photovoltaic power facilities built by the renewable
in other regions (Kumar and Malvoni, 2019). A cost analysis of
energy subsidiary of the state-owned power provider Sonelgaz
solar power systems located in Turkey shows that one of the
(SKTM) (SKTM, 2019). It covers a rectangular area about 1000 m
most important financial factors blocking investment in renew-
long and 400 m wide about 0.3 km east of route CW N◦ 73
able energy sources was high interest rates (Gürtürk, 2019). (27◦ 54′ 15.71′′ N and 0◦ 18′ 45.25′′ W) as shown in Fig. 2. The axis of
The growing scale photovoltaic power plants around the world the site deviates by −58◦ from due south. The altitude is about
and the importance of developing of grid codes for their inte- 258 m. The site comprises 20 photovoltaic subfields of 1 MWp
gration has been discussed (Cabrera-Tobar et al., 2016a). The im- formed by a two-inverter conversion station and a transformer
pact of wind direction on the overall performance of the utility- station. The entrance, command shelter and evacuation outpost
scale photovoltaic plant has been examined using six months are in the east central part of the site. The general layout, equip-
of data collected from the Hadley solar plant in the United King- ment, and operation of the main systems and components, the
dom from January 1 to July 1, 2017, which revealed that power photovoltaic field (solar panels, load-bearing structures, assembly
production increased significantly under southerly wind condi- boxes, cabling, conversion stations and transformer stations are
tions while keeping all other effective factors the same (Vasel described below. Design parameters of the Adrar 20 MVp photo-
and Iakovidis, 2017). An economic analysis of a photovoltaic voltaic power plant are presented in Table 2. The energy produced
and hydrogen turbine hybrid 100 MW power plant found that is intended to increase the reliability of the local electricity grid,
electricity could be provided at $0.12/kWh in the average-case which supplies power to about 27 100 houses.
scenario, and $0.16/kWh in the worst-case scenario, with payback
periods of 13 years and 15 years respectively, based on an 8% 3.3. Description of the subfields and modules
interest rate (Ebaid et al., 2015). Finally, based on analysis of eco-
nomic feasibility using HOMER, interior climatic zones would be To achieve the desired power, the photovoltaic plant archi-
tecture factors in the constraints of compatibility with inverter
preferred for photovoltaic/diesel hybrid systems (Suresh Kumar
characteristics were summarized in Table 3.
and Manoharan, 2014). Sensitivity analysis showed that the net
The panels contain Yingli type solar cells made from blocks of
present cost of such a system can be reduced. It is noteworthy
poly-crystallized silicon. These cells cost less to manufacture than
that worldwide global solar energy capacity reached 714 GW in
monocrystalline cells, but their energy loss is considerable in hot
March 2021 (IRENA, 2021).
climates (Tossa et al., 2016). With 81 840 modules rated 245 Wp,
the installed power generation capacity is 20.0508 MWp or 1.003
3. Methods and materials MWp per subfield of 4092 modules. Each subfield is equipped
with two inverters and a step-up transformer. The inverters (out-
put voltage 315 V AC) are connected to the low-voltage side via
3.1. Site description
AC cables to the 1.250 kVA transformer, which raises the voltage
to 31.5 kV. The 20 subfields are connected to the 30 kV evacuation
African deserts, especially in Algeria, have immense potential station. The electricity produced is evacuated by a 30 kV overhead
for solar energy generation. The practically constant insolation line connected to the local grid as shown in Fig. 3.
amounts to 3500 h of annual sunshine, which is equivalent to
more than 6000 trillion kWh. Almost all parts of Algeria receive 3.3.1. The 1 MW subfields
5–7 kWh of solar radiation per square meter per day (Bentouba Each subfield is equipped with three-level grouping boxes.
and Bourouis, 2016). Table 1 shows the daily insolation in four The photovoltaic modules are connected to 500 kW inverter
Algerian provinces. cabinets through junction boxes (level 1), parallel boxes (level
7299
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Fig. 2. Aerial view of the Adrar solar (photovoltaic) power plant (SKTM, 2019).

Fig. 3. General scheme of the Adrar 20 MW power plant connected to the grid.

2) and general boxes (level 3). Direct current is converted into cables and hence losses therein. A subfield or transformer sub-
alternating current and sent to the transformer station. The three- station failure does not affect production by the other subfields,
stage grouping boxes reduce the total length of the DC cables thus ensuring power deliverability.
and ohmic losses therein and improve the efficiency of power For each 1 MW subfield, a transformer substation of 1.250 kVA
generation throughout the plant as shown in Fig. 4. is planned to include:
- A dry main transformer of 31.5 ± 2 × 2.5%/0.315/0.315 kV.
- Three 31.5 kV cells from the ring circuit to the SF6 isolation.
3.3.2. Transformer substation
The transformer substations of all subfields are connected by 3.4. Photovoltaic modules
31.5 kV medium-voltage cables forming two ring circuits each
connecting 10 substations, and then to the 30 kV evacuation grid. The characteristics of the Yingli 245 W polysilicon photo-
Each ring circuit reduces the length of the medium-voltage AC voltaic solar modules chosen by SKTM are shown in Table 4.
7300
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Fig. 4. Schematic outline of a 1 MW subfield in the Adrar photovoltaic power plant.

Table 4 3.8. Direct current grouping boxes

Power plant panel characteristics (SKTM, 2019).
Module type Yingli Solar YL245-29b The modules are connected via grouping boxes. These three-
Measured power 245 W (±5 W) level direct-current connection boxes are included in the
Measured current 8.28 A
schematic representation of power plant architecture.
Open-circuit voltage 37.7 V
I (continuous current) 8.83 A (DC)
Fire resistance class C 3.8.1. Junction boxes (level 1)
Application class A The 24 level-1 direct-current junction boxes (1 for each 1 MW
Nominal operating cell temperature NOCT ◦ C 46 ± 2
subfield) connect the photovoltaic module chains and transmit
Temperature coefficient of Pmax γ %/◦ C −0.45
the produced power to the parallel boxes (level 2). Each has 8
inputs and 1 output installed on the load-bearing structures of the
panels. They thus each can receive 8 module chains. Each input
3.5. Choice of optimal inclination angle (two polarities) is equipped with a current fuse protection. Each
chain of 22 modules is connected to one junction box, as shown
The angle of inclination of the panels is not adjusted through- in Fig. 4.
out the year. The angle that maximizes the annual irradiation is
VJl1 = NS ∗ Vmpp (1)
27◦ as shown in Fig. 5, which was calculated using HOMER Pro
based on the curve of monthly data. IJl1 = Impp ∗ Np (2)

3.6. The distance between module rows 3.8.2. Parallel boxes (level 2)
The level-2 parallel boxes connect the level-1 junction boxes
The panels must also be spaced optimally to avoid shad- to the level-3 general boxes. A total of 8 parallel DC lightning
ing (Radziemska and Klugmann, 2002). This spacing was deter- protection boxes with 4 inputs and 1 output are installed per
mined again using HOMER Pro software, which considers the subfield. Parallel boxes contain 3 level-1 junction box outputs
daily change in the path of the sun throughout the year, inde- each.
pendently of the exact site location and plant layout. The optimal VJl2 = NS ∗ Vmpp (3)
spacing between the rows of the north–south facing panels based
on the actual field size was 8 m and the distance is the same for IJl2 = Ijl1 ∗ 3 (4)
other photovoltaic power plants installed in this hot region.
3.8.3. General boxes (level 3)
A general box installed at the entrance of each inverter (500
3.7. Main parameters
kW) ensures the connection between the appropriate level-2
parallel box and the solar inverter. Two level-3 general boxes
Two photovoltaic strings (22 panels) are installed per support-
(level 3) with 4 inputs and 4 outputs per subfield are installed
ing structure in a 2 x 11 panel matrix in landscape orientation.
in the conversion station. Each general box contains 4 level-2
The maximal height of the supporting structures is approximately
(parallel box) outputs, as implied in Eqs. (5) and (6).
2.37 m for the panel used in the Adrar power plant with the size
of 1.650 mm x 990 mm x 40 mm. VJl3 = NS ∗ Vmpp (5)
7301
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Fig. 5. Calculation of the optimal angle for solar panels at the Adrar power plant.

IJl3 = Ijl2 ∗ 4 (6) Table 5

Cable dimensions in the Adrar 20 MW power plant.
The general boxes are equipped with four inputs (two polari- Equipment Diameter Length
ties), four outputs (two polarities) up to 240 mm2 and 4 DC fuse Junction box (level 1) 4 mm2 91.5 km
switches for system protection. Parallel box (level 2) ≥ 70 mm2 30.5 km
Each inverter (500 kW) is connected to a level-3 junction General box (level 3) 240 mm2 27.0 km
which comprises 4 level-2 junctions and each of which is con- 3 three-core AC cables for connection
nected to 3 level-1 junctions. Each 1 MW subfield thus comprises between the AC side of the inverter and the
a total of 24 junctions, 6 junction boxes containing 7 chains of 22 low-voltage side of the main transformer.
Aluminum section = 3 ∗ 240 mm2
modules and 18 junction boxes with 8 chains of 22 modules. A
one-megawatt subfield thus comprises 4092 panels. All junctions 31.5 kV single-core cable to the aluminum 3 ∗ 240 mm2 1.8 km
core with XLPE insulation for ring circuits
are equipped with a lightning arrester (two-polarity lightning between transformer stations and the
protection) to protect the electrical system against direct shock evacuation outpost
and inductive shock and with a monitor to control the current and
voltage on each string. These structures are inside the box. Real-
Table 6
time data can be sent via an RS485 serial cable. The box housing
Main characteristics of the inverters (DC ≥ AC converters) (SKTM, 2019).
is made of metal to ensure long-lasting and stable operation in
Inverter Specification
an outdoor environment.
Rated AC power 500 kW
Max. output power 550 kVA
3.9. Cabling Output frequency range 47∼52 Hz
Maximum efficiency 98.5%
The lengths of the cables from the panels to the transformers Max. DC input voltage 1000 V
Max. DC input current 1100 A
(Table 5) are calculated from the three-level single-line diagram
MPPT voltage ∼500–820 (VDC)
of the grouping boxes, the ring circuit connection of the trans- Voltage on cable output 315 (VAC)
former stations and the layout of the photovoltaic fields on the Power factor ∼0.9–1
site. Harmonic distortion rate [%] <IP20
Envelope protection level 3% (nominal power)
Ambient temperature range ∼30–55 ◦ C
3.10. Inverters

The inverter is an electronic device that converts direct current

produced by photovoltaic modules to alternating current using 3.11. Transformers
control and protection circuitry. It can accept the maximal current
and voltage produced by the photovoltaic field. The power plant
The transformer substation is used to raise the output voltage
is equipped with 40 inverters of 500 kW DC/AC, 2 per sub-
of the two inverters and send the electricity generated to the
field. The ∼520–820 VDC input range ensures AC output voltage
stability with a maximal current of 1100 A at high efficiency evacuation outpost. There are 20 transformers in the power plant
(maximum ≥ 98.5%). The DC side of the inverters has 4 two- as one transformer substation for each 1 MWp subfield. SUNTUN
polarity inputs each equipped with direct current fuse protection, manufactured the 1250 kVA class transformers to step up the
a general disconnect switch and a DC lightning arrester. The low-voltage output of the inverters to a medium voltage of 30
technical specifications of the inverters are shown in Table 6. kV.
7302
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

3.12. Photovoltaic penetration sun hours or insolation in units of kWh/m2 and defines the
solar radiation resource potentially available to the system. It
The cumulative photovoltaic power generation capacity in- is a function of photovoltaic array location and orientation and
stalled in Algeria was 448 MW at the end of 2019 (IRENA, 2021), month-to-month and year-to-year weather variability (Marion
which was still less than 3% of peak demand. On July 12th 2021, et al., 2005):
at 14:30, a new historic peak in power consumption estimated ∫ t2
(solar radiation) dt
at 16065 MW was recorded https://www.radioalgerie.dz/news/fr/ Yr = (8)
article/20210713/214632.html, http://www.aps.dz/economie/915 t1 1000
71-electricite-un-nouveau-record-de-consommation. In its quest where t1 and t2 are the instantaneous start and end times for each
to meet the growing demand for electricity using sustainable data record and the integral gives the in-plane irradiation of the
resources, Algeria could follow the example of its southern region, array over that time interval.
which now has a micro-grid network comprising 7 photovoltaic
H
power plants each generating 33 MW, one wind farm generating Yr = (9)
10.2 MW, albeit more than 500 MW generated by fossil fuels Gref
(Table 7). Photovoltaic electricity now powers 27 100 houses in H is the derived array plane insolation in Wh/m2.
the region, plus hospitals, schools and a university. Gref is the reference irradiance at STC (1000 W/m2 )

4. Methodology 4.3. Performance ratio

Twenty-two months of energy production data were obtained The performance ratio indicates the degree of utilization of
from the AC billing meters provided by the distribution system a photovoltaic system as a whole (International Electrotechnical
operator SKTM (renewable energy subsidiary of the state-owned Commission, 1998). It is the ratio of the normalized parameters
power provider Sonelgaz), which logs data every 2 min. The Yf and Yr , that is, the solar energy converted and utilized as AC
uncertainty of the AC billing meters is small (0.2–0.5%). The data current divided by the energy that could have been generated
combine production from all large photovoltaic power plants under ideal conditions of operation and utilization (Padmavathi
built in Adrar. The analysis and conclusion from this large and and Daniel, 2013).
optimized facility can be considered as a reference for calibrating Yf
new projects to be built in similar hot climatic regions of Africa PR = (10)
Yr
and MENA. It is important to analyze data from photovoltaic
power plants that inject electricity into the grid rather than with 4.4. Capacity factor
installed capacity.
The parameters of solar energy systems and components The capacity factor is defined as the ratio of the actual energy
thereof have been established by the International Energy Agency output in AC current to the amount of energy that the system
(IEA) Photovoltaic Power Systems Program and are described in would generate if it operated at nominal power continuously
IEC standard 61724 (Şenol et al., 2016). Those most relevant to throughout the year (for 8760 h).
our analysis are energy output, array yield, final yield, reference
actual energy output EAC
yield, module efficiency, inverter efficiency, system efficiency, CF = = × 100 (11)
energy loss (array capture loss and system loss), performance 8760 · system rated power Pnom × 24
ratio and capacity factor. These normalized indicators serve as key The capacity factor represents the energy delivered by an
comparators for evaluating the performance of this large-scale electrical power generating system, whether conventional or
photovoltaic power plant (Marion et al., 2005; Al-Badi, 2018; renewable-resource-based (ElhadjSidi et al., 2016).
Padmavathi and Daniel, 2013).
EDC
These performance parameters allow the detection of oper- Photovoltaic efficiency, τpv = (12)
ational problems, which facilitate the comparison of systems. PV area ∗ irradiation
inverter export
However, that may differ concerning design, technology, or ge- Inverter efficiency, τIn = (13)
ographic location, and validate models for system performance EDC
estimation during the design phase. EAC
System efficiency, τs = (14)
PV area ∗ irradiation
4.1. Final yield rating (Yf ) The units of the terms EDC , inverter export and EAC are kWh.
The units of irradiation are kWh/m2 . The capacity factor is tech-
The final yield rating is the energy produced by the field nically more suitable for power plants that run continuously. Like
divided by the peak power of the array. It represents the number most renewable energy systems, solar photovoltaic power plants
of hours that the array would need to operate at its rated power run intermittently, since insolation is never constant or even
to provide the same energy as at peak power (Marion et al., 2005). available for 24 h. The CF ranges from 0.05 to 0.30 for photovoltaic
power plants. This performance measurement is unavoidable in
actual output AC energy EAC comparison with conventional power generation.
Yf = = (hours) (7)
rated array capacity Prated
5. Results and analysis
Prated is the rated capacity of the array in kW.
EAC is actual array output energy in kWh. 5.1. Data monitoring system

4.2. Reference yield Yr The performance of the photovoltaic plant is assessed using
a huge database logged via a sensor-based system that monitors
The reference yield is the total in-plane irradiance divided the electrical status of the panels, junction boxes, inverters, and
by the reference irradiance. It represents the number of peak transformers as well as the local weather conditions. The SCADA
7303
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Table 7
Topology of the electrical micro-grid in the Adrar/In Salah/Timimoun region of Algeria (SKTM, 2019).
Province Fossil-fuel power plants (number × MW) Total MW Renewable energy power plants
Adrar 4 × 25 + 2 × 8 116 20 MW photovoltaic
In Saleh 4 × 24 96 6 photovoltaic
Kabertan 2 × 20 40 10.2 MW wind + 3 MW photovoltaic
Timimoun 2 × 20 + 2 × 25 90 9 MW photovoltaic
Zaouit Kounta 8 × 20 160 6 MW photovoltaic
Reggane 0 0 6 MW photovoltaic
Aoulef 0 0 6 MW photovoltaic
Total 24 502 66.2 MW photovoltaic+wind

Fig. 6. Solar radiation, monitored on the ground and calculated from satellite
data. Fig. 7. Wind speed, monitored by weather stations and calculated from satellite
data.

control command system links a central control room to the

power plant, annexed installations, and the 30 kV substation. In
addition to monitoring and control, the system plays a central
role in communication. Any equipment operation that is not
under its authority is still monitored, making all gatherable in-
formation available on its network. The power plant runs all the
time except when stopped to solve technical problems. Global
horizontal irradiance (GHI), direct normal irradiance (DNI), am-
bient temperature (Ta ), DC and AC electrical energy generated
(EDC and EAC ) are measured instantaneously at 2-minute intervals.
Wireless transmitters are used to log data from wind speed and
relative humidity sensors.

5.2. Meteorological data

Irradiance, temperature and wind speed data were collected

from November 1, 2016 through December 31, 2018 (two years
and two months) via SCADA. RETScreen weather data were pro-
duced by Environment Canada and HOMER Pro by NASA. The Fig. 8. Temperature, monitored by weather stations and calculated from satellite
data.
monthly variation of average daily insolation on the panels (mon-
itoring, RETScreen Expert and HOMER Pro) is shown in Fig. 6.
Real data and software data differed somewhat. The measured
average insolation varied from 5.422 kWh/m2 /d in December and software calculation give differing profiles. In these cases, the
to 7.34 kWh/m2 /d in September compared to 3.36 kWh/m2 /d real weather data were obtained from local weather stations near
and 7.82 kWh/m2 /d (July) for HOMER Pro and 3.83 kWh/m2 /d the power plant.
to 8.14 kWh/m2 /d (June) for RETScreen Expert. HOMER Pro and The average daily ambient temperature measured by the SKTM
RETScreen use algorithms to calculate insolation from satellite weather station in the power plant varied from 19.1 ◦ C in January
data. In any case, insolation was found to be consistently higher to 43.4 ◦ C in July. During the monitored period, module tempera-
in the months of March through August than September through tures reached 69.73 ◦ C at an irradiance of 1024 W/m2 and a wind
February. It was greater than 7 kWh/m2 per day for the six speed of 5.7 m/s. Wind over the modules lowers the cell operating
months of Spring and Summer. temperature. The average wind speed varied from 5.63 m/s in
The comparisons of wind speed and of temperature data from February to 6.11 m/s in July (Fig. 7). Although the profiles are the
the three sources are shown in Figs. 7 and 8. Again, monitoring same, HOMER Pro disagrees considerably with RETScreen and the
7304
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Table 8
Analysis of power plant performance for the year 2017 (based on real data).
Month (2017) Yf Performance ratio Capacity factor
1 4.847903 78.44504 20.1996
2 4.619643 70.20734 19.24851
3 5.000194 67.7533 20.83414
4 5.296867 72.9596 22.07028
5 5.21129 73.29522 21.71371
6 5.029133 69.4814 20.95472
7 4.652387 64.02691 19.38495
8 5.125806 69.73981 21.35753
9 4.94025 70.92863 20.58438
10 5.444274 84.3054 22.68448
11 5.01975 84.40527 20.91563
12 4.977581 91.80849 20.73992

5.4. The influence of temperature

As shown in Figs. 14 and 15, the temperature measured in

Fig. 9. Monthly electricity production, monitored and calculated by HOMER Pro
and RETScreen Expert. the installation lags daily power output, especially in the winter.
Despite the long hours of insolation, the power loss is huge in
the summer because of the climatic conditions and the choice of
real data. Air temperature is correlated weakly with electricity polycrystalline silicon photoelectric cells.
production.
5.5. HOMER Pro and RETScreen system configuration
5.3. Electricity production
HOMER Pro defines simulation parameters in more detail than
The monthly production of electricity is shown in Fig. 9.
RETScreen Expert. In our case study, it was close to actual mon-
Monitoring shows 2725.72 MW in July versus 3442.875 MWh
itoring data. In addition, it simulates a viable system for all
in March. HOMER Pro calculated 2382.54 MWh in December
possible combinations of necessary equipment considering the
and 3130.905 MW in May and RETScreen Expert gave 3066.45
case where no electrical energy storage is planned like our 20
MWh in December and 3969.52 MWh in March. Annual pro-
MW photovoltaic power plant. HOMER simulates the operation
duction recorded by monitoring was 36,363.875 MWh versus
34 241.24 MWh calculated by HOMER Pro and 42 338.57 MWh of the photovoltaic power plant for a whole year, in time steps
by RETScreen Expert. HOMER Pro is thus closer to reality. In all from one minute to one hour. All the electricity produced is sold
three cases, the average daily energy output was higher in March to the grid, as shown in Fig. 16. In general, HOMER Pro can
through November than in January through February. model photovoltaic systems that combine any type of conven-
Electrical energy output was essentially proportional to daily tional energy, renewable energy, storage, and load management,
average insolation, as shown in Fig. 10 for monitoring data and making decisions reliable, so that any system can be designed
in Fig. 11 for HOMER Pro. consistently.
The linearity of the relationship between insolation and elec- HOMER Pro calculations of yearly solar radiation, grid sales,
tricity production is apparent when a winter day is compared to inverter input, inverter output, power output, ambient and cell
a summer day (Figs. 12 and 13). RETScreen Expert could not be temperature, and total renewable energy output are represented
compared since it generated only monthly and yearly data. in Figs. 17 and 18.

Fig. 10. Monitored daily production of electricity versus insolation during winter days.

7305
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Fig. 11. HOMER-pro-calculated daily production of electricity versus insolation during winter days.

Fig. 12. Monitored daily production of electricity versus insolation during summer days.

5.6. Performance analysis Table 9

Analysis of power plant performance for the year 2018 (based on real data).
Month (2018) Yf Performance ratio Capacity factor
The performance of the power plant is assessed using the AC
performance ratio, the overall system efficiency, and the capacity 1 4.895194 79.21025 20.39664
2 5.003857 76.04646 20.8494
factor. In addition, the various capture loss parameters such as 3 5.553024 75.24423 23.1376
temperature provide valuable information about the magnitude 4 5.41175 74.54201 22.54896
of losses (see Tables 8 and 9). 5 5.22629 73.50619 21.77621
Fig. 19 shows the monthly variation of the electrical en- 6 4.888833 67.54305 20.37014
7 4.396323 60.5029 18.31801
ergy delivery capacity factor as determined from the power 8 4.75029 64.63068 19.79288
plant monitoring system and is calculated by the software. The 9 4.681875 67.21906 19.50781
HOMER Pro calculations track the real data quite closely except 10 4.818629 74.61719 20.07762
for November, December, and January. 11 4.89675 82.33707 20.40313
12 5.254105 96.90881 21.8921
The average daily yield is directly proportional to the average
daily insolation, varying from 4.3 h/d in January to 5.7 h/d in
March based on measurement and from 4.9 h/day in January to
5.4 h/day in April based on a calculation by HOMER Pro as shown trend of the reference yield Yr. For all months of the monitoring
in Fig. 20. For most of the year, RETScreen Expert overestimated it, period, there is a difference between average reference yield
especially in the summer. The yield was proportional to electricity and system yield. This difference was expected, due to DC/AC
production and the daily average array yield and followed the conversion losses produced in the inverter.
7306
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Fig. 13. HOMER-pro-calculated daily production of electricity versus insolation during summer days.

Fig. 14. Photovoltaic power output and ambient temperature measured on a typical summer day.

As predictors of the performance ratio (Fig. 21), HOMER Pro is to temperature effects. Inverter efficiency varied from 94% in
quite accurate except in the months of October, November, and July to 98.23% in February.
December, whereas RETScreen Expert consistently overestimated
the ratio, except in these three months. The information obtained 5.7. Influence of temperature on the photovoltaic system
from the simulation can be useful for investigating the charac-
teristics of the photovoltaic system under different conditions To explain the influence of temperature on module (and hence
(temperature, insolation) or the impact of cell technology. Energy power plant) efficiency, we used RETScreen and HOMER Pro
to simulate the power capacity of a 20 MW power plant built
generation predicted by simulation gives an idea of power plant
with micro-crystalline or thin-film solar cells. Thin-film cells re-
performance when precise measurements of radiation, power
portedly perform well in the summer compared to crystalline
and energy exported under varying environmental conditions
silicon cells due to their low temperature coefficient (Elhad-
are available. Several physical and statistical models have been
jSidi et al., 2016). The results of RETScreen simulations with
established to predict the output and overall performance of poly-crystalline silicon cells (Yingli Solar YL254-29b) replaced by
photovoltaic systems (Tsoutsos et al., 2005; Tian et al., 2009). The mono-crystalline (Canadian Solar CSK 56 295 MS) or thin film
results shown here should guide researchers and engineers who (First Solar series 4 107 Cd/Te - FS-4105 A) cells under the same
are modeling and planning solar photovoltaic power projects. conditions are shown in Table 10. The thin-film cells are clearly
Sizing is a crucial aspect of solar power plant design. Modeling expected to perform better than either crystalline silicon solar cell
focuses not only on sizing but also on optimization for operation in hot climates, which has been confirmed elsewhere (ElhadjSidi
under variable meteorological conditions. et al., 2016). In addition, the capital cost of a large power plant us-
In Fig. 22, the monthly average photovoltaic system efficiency ing thin-film cells would be lower than one using poly-crystalline
for 2018 varied from 10.34% in July to 15.04% in November, due silicon cells.
7307
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Fig. 15. Photovoltaic power output and ambient temperature measured on a typical summer day.

Table 10
Comparison of monthly photovoltaic electrical energy generation (MWh) simulated on HOMER Pro and RETScreen Expert for a 20 MW power plant operating in a
hot climate.
Month RETScreen Expert HOMER Pro
Thin filma Mono-crystallineb Poly-crystallinec Thin film Mono-crystalline Poly-crystalline
January 3535.8934 4080.0128 3480.096 2832.866 2630.407 2824.766
February 3362.9426 4386.1853 3285.161 2997.656 2765.494 2982.976
March 4096.0854 4286.1243 3969.518 3669.913 3354.576 3641.157
April 3896.1797 4100.2547 3751.558 3548.235 3219.980 3510.405
May 3902.9448 4032.3182 3732.318 3471.689 3130.906 3425.528
June 3790.4596 4085.875 3585.228 3172.108 2840.194 3121.276
July 3904.7411 2905.8547 3673.968 3401.284 3026.438 3341.015
August 3949.029 3014.6854 3715.633 3416.137 3038.213 3356.639
September 3660.6918 3256.3654 3472.225 3278.766 2931.863 3228.856
October 3574.3383 3149.2001 3437.677 2951.164 2685.214 2920.978
November 3248.4903 3358.2014 3167.7504 2693.460 2476.167 2676.271
December 3110.1138 3724.5684 3066.4521 2591.693 2408.362 2583.699
Total 44 031.91 44 132,82 42 337.5855 38 024.972 34 507.814 37 613.569
a
First Solar series 4 107 CdTe - FS-4105A.
b
Mono-crystalline Canadian Solar CSK 56 295 MS.
c
Yingli Solar YL254-29b.

where Tcell , Tamb are respectively the cell and ambient temper-
atures, H is irradiance in W/m2 , and Tnco is the manufacturer-
specified nominal cell operating temperature (45 ◦ C).
Monthly average ambient temperatures varied from 19.1 ◦ C
in January to 43.4 ◦ C in July. The corresponding monthly average
cell temperatures were 44.1 ◦ C and 68.4 ◦ C. The curves are not
perfectly symmetrical and are highly coherent with each other
(Fig. 23).
The lower efficiencies measured in June, July and August can
be attributed to temperature effects and solar technologies as
indicated in Fig. 23 and Table 10. The negative slope of the
efficiency versus temperature regression indicates that higher
Fig. 16. System configuration as modeled by HOMER Pro.
ambient temperature has an adverse effect on overall efficiency.
Real data and HOMER Pro calculations differ considerably
(Fig. 24). The software appears to underestimate considerably
5.7.1. Effect of temperature on photovoltaic cell efficiency the influences of temperature and panel technology. On a hot
day such as July 27, 2018, the actual production was 6000 kW,
Ambient temperature and cell temperature display a linear re-
while the HOMER Pro estimate was 14,000 kW. The temperature
lationship. The increase in module temperature causes the elec- has a stronger effect on the efficiency of the Yingli YL254-29b
polycrystalline silicon solar panel than was attributed in the
trical power output to decrease (Yao et al., 2014).
software, which appears to be quite indifferent to high-precision
meteorological data. Actual energy loss is more than 50% dur-
Tcell = Tamb + H × (Tnco − 25) /800 (15) ing peak hours on days when the panel temperature exceeds
7308
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Fig. 17. HOMER Pro output results.

50 ◦ C for the whole day, decreasing the power plant output. whereas temperature and cell technology affect power produc-
In some cases, for small application, PV/T system was indicated
tion on longer time scales. Specific molecules and aerosols in air
as one of the best ways to reduce the dependence of PV on
temperature (Atmaca and Pektemir, 2019). also have an influence. Since the orientation and inclination of
the photovoltaic panels are known, it is possible to forecast the
5.8. Photovoltaic forecasting and key indicators
peak power output on clear sky days. We used statistical analysis
Models that use forecasted irradiance like those that use fore-
to validate photovoltaic forecasting by comparing the predictions
casted wind to predict power output are gaining in popular-
ity. Clouds and fog have the biggest influence on insolation, with real data obtained by direct monitoring.
7309
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Fig. 18. Output power results predicted by HOMER Pro.

5.8.1. Statistical analysis of power output forecasting 2013). The t statistic allows us to state with a specified level
The statistical indicators used to compare model results to real of confidence whether the model predictions are accurate or
data were the mean bias error (MBE), root mean square error not (Sundaram et al., 2015; Yao et al., 2014).
(RMSE), mean percentage error (MPE) and the t statistic as de-
fined in the literature (Tarhan and Sarı, 2005). The MBE provides 5.8.2. Root mean square error (RMSE)
an assessment of the long-term performance of the simulation. A The root mean square error can be computed using the follow-
positive value indicates overestimation, a negative value indicates ing equation. Its value is always positive, zero meaning perfect
under-estimation by the simulation software, the smaller the prediction accuracy (never achieved in practice) (Stone, 1993;
absolute value, the greater the exactness. The RMSE provides an Padmavathi and Daniel, 2013).
assessment of the short-term performance of the proposed model

 n
by allowing a term-by-term comparison of the actual deviation 1 ∑
RMSE = √ (ci − mi )2 (16)
between predicted and measured values. The MBE and RMSE both n
i=1
provide a rational basis for comparison, but no objective indicator
of the statistical significance of deviations of model predictions Ci Energy capacity installed.
from actual measurements (Stone, 1993; Padmavathi and Daniel, mi Energy measured.
7310
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Fig. 19. Capacity factor during the year 2018, based on real data, HOMER Pro
Fig. 22. Power plant photovoltaic and inverter efficiency.
and RETScreen Expert.

Fig. 23. Ambient temperature and photovoltaic cell temperature (monthly

Fig. 20. Yf during the year 2018, based on real data, HOMER Pro and RETScreen averages).
Expert.

Fig. 21. Performance ratio during the year 2018, based on real data, HOMER
Pro and RETScreen Expert.
Fig. 24. Influence of temperature on electricity production.

7311
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Table 11 6. Conclusion
Accuracy of HOMER Pro and RETScreen Expert at predicting power output of
hot climate photovoltaic power plant.
Software RMSE MPE (%) t statistic MBE The performance of a grid-connected 20 MW photovoltaic
HOMER Pro 351.1 −0.00095 3.75 −176.88
power plant operating in a hot climate was characterized in terms
RETScreen Expert 566.7 0.0016 6.12 497.80 of the influence of insolation, air temperature, solar cell technol-
ogy of electricity production and yield. The performance of the
power plant depended heavily on insolation and environmental
conditions, especially temperature. The solar cell technology used
5.8.3. Mean bias error (MBE)
likely contributed to underperformance. Real data showed that
The mean bias error is expressed as follows (Stone, 1993;
high temperatures negatively affect the behavior of polycrys-
Padmavathi and Daniel, 2013)
talline silicon cells resulting in non-optimal operation in the
n
1∑ Sahara Desert as the cell temperatures reach more than 60 ◦ C
MBE = (ci − mi ) (17) during the hottest months and the electrical production capac-
n
i=1 ity is reduced by more than 40%. Commercial software predicts
An MBE of zero would indicate that all values calculated by the better performance from thin film solar cells under hot climate
model were identical to the measured values throughout the conditions. The power plant could be made at least 5% more
period of study, which never occurs in practice (Oliveira et al., efficient with thin-film solar cells, which are less sensitive to
2002). temperature changes and cost less initially and are therefore
more recommendable for photovoltaic power generation in hot
5.8.4. The t statistic method climates like in southern Algeria and the MENA region in general.
The statistical significance of the deviations of the predicted Commercial simulation tools are very useful for designing and
values from the measured values, that is, the level of confidence sizing large-scale photovoltaic power projects. The impact of
with which it can be said that such deviations are systematic temperature effects, insolation, wind velocity and cell technology
and a characteristic of the comparison under study (Stone, 1993), can be assessed by numerical simulations such as HOMER Pro and
is evaluated using the t statistic (Oliveira et al., 2002). This can RETScreen Expert. HOMER Pro results were found very close to
be computed using both the RMSE and MBE and considers the those recorded by real-time monitoring to predict total produc-
dispersion of the results, which is neglected when these error tion of electricity. In our analysis of case study, HOMER Pro is very
calculations are used separately. The smaller the t value, the more strong by making it easy to compare thousands of possibilities,
accurate are the predictions. investigating all potential combinations of system types and then
√ categorizing the systems according to the selected optimization
(n − 1) MBE2 variable. This gives us the possibility to quantify the impact of
t − stat = (18)
RMSE2 − MBE2 variables other than our control, such as PV technologies solar cell
impact, temperature, wind speed, solar radiation, and understand
5.8.5. Mean percentage error (MPE) how the optimal system changes with these variations like in
The mean percentage error is the percent deviation of the es- hot climates. Therefore, the total electrical energy provided by
timated monthly average daily energy values from the measured this hot climate power plant in Adrar region was 36363.88 MWh
values. The equation is written as follows: in 2018, only about 5.1% more than the amount predicted by
n
HOMER Pro (34 507.81 MWh) and 14% less than what was pre-
dicted by RETScreen Expert (42 338.57 MWh). This underlines the
∑
MPE = E /n (19)
performance of that Homer Pro in this specific case of hot climate.
i=1
The errors in the predictions are due mainly to the inaccuracies
The best regression values generated by RETScreen and HOMER of the weather databases used by the simulators, especially for
PRO as predictors of the real-time performance of the 20 MW the hot summer months. Based on the t statistic, HOMER Pro
power plant are compared in Table 11. was the better of the two simulators, although neither was very
Based on the t statistic, HOMER Pro was closer to the moni- accurate. This analysis of the performance of the hot climate
tored values, but both simulations deviated from reality, by 14% power plant suggests major design errors to be avoided in similar
for RETScreen and by 5% for HOMER PRO, both validated by the photovoltaic power projects in MENA regions and in hot climates
RMSE, MPE and MBE (Ma et al., 2021). HOMER Pro underesti- anywhere in the world. It also shows that large-scale photovoltaic
mated the annual production of electricity (MBE = −176.88) and power plants are feasible and very competitive in such regions
RETScreen overestimated it (MBE = 497.80). These deviations
and that renewable energy programs initiated by MENA countries
between the actual data and the simulation results could be
are justified. These results show that the performance of such sys-
caused factors associated with the Adrar region such as dust accu-
tems does not depend exclusively on insolation but also on solar
mulation (Enaganti et al., 2022), very hot desert climate most of
panel technology, temperature, and environmental conditions of
the year, degradation of photovoltaic modules and inappropriate
operation.
choice of solar cell technology. Degradation observed in fielded
PV modules is higher in hot climatic zones, specifically in hot Nomenclature
and dry zone, as compared to other climates (Omar et al., 2020;
Bansal et al., 2021). The performance of a 2000 MW solar PV
plant operating under the weather conditions in Kuwait, which is PV Photovoltaic
close to the weather of Adrar region, was simulated using Monte HOMER Pro Hybrid Optimization Model for Electric
Carlo approach. The results showed, on average, that power gen- Renewables.
eration was 13% lower in the summer period compared with RETScreen Energetic software developed by resources Canada.
the spring when the temperature is milder and solar production GIS Geographic Information System
peaks (Alshawaf et al., 2020).
7312
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

References
GIPV Grid Interactive PV
GHG Greenhouse Gas Emissions Al-Badi, A.H., 2018. Measured performance evaluation of a 1.4 kW grid connected
IEC International Electrotechnical Commission desert type PV in Oman. Energy Sustain. Dev. 47, 107–113.
IRENA International Renewable Energy Agency Algeria Ministry of Energy. 2021. http://www.energy.gov.dz/francais.
Alshawaf, M., Poudineh, R., Alhajeri, N.S., 2020. Solar PV in Kuwait: The effect of
SCADA Supervisory control and data acquisition. ambient temperature and sandstorms on output variability and uncertainty.
CIS Copper Indium Selenide Renew. Sustain. Energy Rev. 134.
CdTe Cadmium Telluride Atmaca, M., Pektemir, İ.Z., 2019. An investigation on the effect of the total effi-
C-Si Crystalline Silicon ciency of water and air used together as a working fluid in the photovoltaic
thermal systems. Processes 7.
GHG Greenhouse gas Bansal, N., Pany, P., Singh, G., 2021. Visual degradation and performance
SKTM Society Karaba Takkat Moutajadeda (Company of evaluation of utility scale solar photovoltaic power plant in hot and dry
renewable energies). climate in western India. Case Stud. Therm. Eng. 26.
NOCT Nominal operating cell temperature Bentouba, S., Bourouis, M., 2016. Feasibility study of a wind–photovoltaic hybrid
power generation system for a remote area in the extreme south of Algeria.
MPPT Maximum power point tracker
Appl. Therm. Eng. 99, 713–719.
MENA Middle East and North Africa Besarati, S.M., Padilla, R.V., Goswami, D.Y., Stefanakos, E., 2013. The potential of
Yf Final yield Rating. harnessing solar radiation in Iran: Generating solar maps and viability study
Yr Reference yield. of PV power plants. Renew. Energy 53, 193–199.
Cabrera-Tobar, A., Bullich-Massagué, E., Aragüés-Peñalba, M., Gomis-Bellmunt, O.,
Prated Rated power
2016a. Review of advanced grid requirements for the integration of large-
EAC Actual array output energy scale photovoltaic power plants in the transmission system. Renew. Sustain.
PR Performance ratio Energy Rev. 62, 971–987.
CP Capacity factor. Cabrera-Tobar, A., Bullich-Massagué, E., Aragüés-Peñalba, M., Gomis-Bellmunt, O.,
Tcell Cell temperature. 2016b. Topologies for large scale photovoltaic power plants. Renew. Sustain.
Energy Rev. 59, 309–319.
Tamb Ambient temperature. Cazzaniga, R., Cicu, M., Rosa-Clot, M., Rosa-Clot, P., Ventura, C., 2018. Floating
Tnco Manufacturer-specified nominal cell operating photovoltaic plants: Performance analysis and design solutions. Renew.
temperature. Sustain. Energy Rev. 81, 1730–1741.
H Derived array plane insolation in Wh/m2 Ebaid, M.S.Y., Hammad, M., Alghamdi, T., 2015. Thermo economic analysis of PV
and hydrogen gas turbine hybrid power plant of 100 MW power output. Int.
Gref Reference irradiance at STC (1000 W/m2 ) J. Hydrogen Energy 40, 12120–12143.
EDC Energy delivered pb PV on DC ElhadjSidi, C.E., Ndiaye, M.L., El Bah, M., Mbodji, A., Ndiaye, P.A., 2016. Perfor-
GHI Global horizontal irradiance mance analysis of the first large-scale (15 MWp) grid-connected photovoltaic
DNI Direct normal irradiance plant in Mauritania. Energy Convers. Manage. 119, 411–421.
Enaganti, K.P., Bhattacharjee, A., Ghosh, A., Chanchangi, N.Y., Chakraborty, C.,
RMSE Root mean square error. Mallick, T.K., Goel, S., 2022. Experimental investigations for dust build-up
MPE Mean percentage error. on low-iron glass exterior and its effects on the performance of solar PV
MBE Mean bias error. systems. Energy 239.
t-stat t statistic. Guerrero-Lemus, R., Cañadillas-Ramallo, D., Reindl, T., Valle-Feijóo, J.M., 2019.
A simple big data methodology and analysis of the specific yield of all PV
power plants in a power system over a long time period. Renew. Sustain.
Energy Rev. 107, 123–132.
CRediT authorship contribution statement
Gürtürk, M., 2019. Economic feasibility of solar power plants based on PV
module with levelized cost analysis. Energy 171, 866–878.
Hammad, B., Al-Abed, M., Al-Ghandoor, A., Al-Sardeah, A., Al-Bashir, A., 2018.
Said Bentouba: Conception and design of study, Acquisition
Modeling and analysis of dust and temperature effects on photovoltaic
of data, Analysis and/or interpretation of data, Writing - original systems performance and optimal cleaning frequency: Jordan case study.
draft, Writing - review & editing. Mahmoud Bourouis: Concep- Renew. Sustain. Energy Rev. 82, 2218–2234.
tion and design of study, Analysis and/or interpretation of data, IRENA, 2021. Renewable Capacity Statistics 2021. International Renewable
Energy Agency (IRENA), Abu Dhabi.
Writing - original draft, Writing - review & editing. Nadjet Zioui: Kumar, N.M., Gupta, R.P., Mathew, M., Jayakumar, A., Singh, N.K., 2019. Perfor-
Conception and design of study, Analysis and/or interpretation of mance, energy loss, and degradation prediction of roof-integrated crystalline
data, Writing - original draft, Writing - review & editing. Aru- solar PV system installed in Northern India. Case Stud. Therm. Eng. 13,
100409.
mugam Pirashanthan: Writing - original draft, Writing - original Kumar, M., Kumar, T.A., 2017. Performance assessment and degradation analysis
draft. Dhayalan Velauthapillai: Conception and design of study, of solar photovoltaic technologies: A review. Renew. Sustain. Energy Rev. 78,
Writing - original draft, Writing - review & editing. 554–587.
Kumar, N.M., Malvoni, M., 2019. A preliminary study of the degradation of large-
scale c-Si photovoltaic system under four years of operation in semi-arid
climates. Results Phys. 12, 1395–1397.
Declaration of competing interest Kumar, B.S., Sudhakar, K., 2015. Performance evaluation of 10 MW grid
connected solar photovoltaic power plant in India. Energy Rep. 1, 184–192.
Ma, Y., Lv, Q., Zhang, R., Zhang, Z., Zhu, H., Yin, W., 2021. Short-term photo-
The authors declare that they have no known competing finan- voltaic power forecasting method based on irradiance correction and error
cial interests or personal relationships that could have appeared forecasting. Energy Rep. 7, 5495–5509.
to influence the work reported in this paper. Marion, B., Adelstein, J., Boyle, K., Hayden, H., Hammond, B., Fletcher, T.,
Canada, B., Narang, D., Kimber, A., Mitchell, L., Rich, G., Townsend, T.,
2005. Performance parameters for grid-connected PV systems. In: Conference
Record of the Thirty-First IEEE Photovoltaic Specialists Conference.
Acknowledgments Martín-Martínez, S., Cañas-Carretón, M., Honrubia-Escribano, A., Gómez-
Lázaro, E., 2019. Performance evaluation of large solar photovoltaic power
plants in Spain. Energy Convers. Manage. 183, 515–528.
The authors gratefully acknowledge the Algerian Company of
Necaibia, A., Bouraiou, A., Ziane, A., Sahouane, N., Mouhadjer, S., 2018. Ana-
Electricity and Renewable Energy SKTM (Shariket Kahraba wa lytical assessment of the outdoor performance and efficiency of grid-tied
Taket Moutadjadida) for collaborating in this investigation and photovoltaic system under hot dry climate in the south of Algeria. Energy
the supply of design and real monitoring data of the power Convers. Manage. 171, 778–786.
Oliveira, A.P., Escobedo, J.F., Machado, A.J., Soares, J., 2002. Correlation models of
plant. All authors approved the version of the manuscript to be diffuse solar radiation applied to the city of São Paulo, Brazil. Appl. Energy
published. 71, 59–73.

7313
S. Bentouba, M. Bourouis, N. Zioui et al. Energy Reports 7 (2021) 7297–7314

Omar, N.I., Boukhattem, L., Oudrhiri, H.F., Bennouna, A., Oukennou, A., 2020. Out- SKTM, 2019. Shariket Kahrabawa Taket Moutadjadida. http://www.sktm.dz/.
door performance analysis of different PV technologies under hot semi-arid Stone, R.J., 1993. Improved statistical procedure for the evaluation of solar
climate. Energy Rep. 6 (6), 36–48. radiation estimation models. Sol. Energy 51, 289–291.
Padmavathi, K., Daniel, S.A., 2013. Performance analysis of a 3MWp grid Sundaram, S., Sarat, J., Babu, C., 2015. Performance evaluation and validation
connected solar photovoltaic power plant in India. Energy Sustain. Dev. 17, of 5MWp grid connected solar photovoltaic plant in South India. Energy
615–625. Convers. Manage. 100, 429–439.
Pavlović, T., Milosavljević, D., Radonjić, I., Pantić, L., Radivojević, A., Pavlović, M., Suresh Kumar, U., Manoharan, P.S., 2014. Economic analysis of hybrid power sys-
2013. Possibility of electricity generation using PV solar plants in Serbia. tems (PV/diesel) in different climatic zones of Tamil Nadu. Energy Convers.
Renew. Sustain. Energy Rev. 20, 201–218. Manage. 80, 469–476.
Pothecary, S., 2016. Solar to account for 30% of all generation capacity Susanto, J., Shahnia, F., Ludwig, D., 2018. A framework to technically evaluate
investment until 2040 – PV Magazine International. https://www.pv- integration of utility-scale photovoltaic plants to weak power distribution
magazine.com/2016/06/13/solar-to-account-for-30-of-all-generation- systems. Appl. Energy 231, 207–221.
capacity-investment-until-2040_100024957/. Tarhan, S., Sarı, A., 2005. Model selection for global and diffuse radiation over
Purohit, I., Purohit, P., 2018. Performance assessment of grid-interactive solar the Central Black Sea (CBS) region of Turkey. Energy Convers. Manage. 46,
photovoltaic projects under India’s national solar mission. Appl. Energy 222, 605–613.
25–41. Tian, Y., Zhao, L., Meng, H., Sun, L., Yan, J., 2009. Estimation of un-used land
Radziemska, E., Klugmann, E., 2002. Thermally affected parameters of the potential for biofuels development in (the) People’s Republic of China. Appl.
current–voltage characteristics of silicon photocell. Energy Convers. Manage. Energy 86, 77–85.
43 (14), 1889–1900. Tossa, Alain K., Soro, Y.M., Thiaw, L., Azoumah, Y., Sicot, Lionel, Yamegueu, D.,
Rashwan, S.S., Shaaban, A.M., Al-Suliman, F., 2017. A comparative study of a Lishou, Claude, Coulibaly, Y., Razongles, Guillaume, 2016. Energy perfor-
small-scale solar PV power plant in Saudi Arabia. Renew. Sustain. Energy mance of different silicon photovoltaic technologies under hot and harsh
Rev. 80, 313–318. climate. Energy 103, 261–270.
Rehman, S., Ahmed, M.A., Mohamed, M.H., Al-Sulaiman, F.A., 2017. Feasibility Tsoutsos, T., Frantzeskaki, N., Gekas, V., 2005. Environmental impacts from the
study of the grid connected 10MW installed capacity PV power plants in solar energy technologies. Energy Policy 33, 289–296.
Saudi Arabia. Renew. Sustain. Energy Rev. 80, 319–329. Vasel, A., Iakovidis, F., 2017. The effect of wind direction on the performance of
Satsangi, K.P., Bhagwan, D.D., Sailesh. A.K. Saxena, B.G.S., 2018. Performance solar PV plants. Energy Convers. Manage. 153, 455–461.
analysis of grid interactive solar photovoltaic plant in India. Energy Sustain. Yadav, A.K., Sharma, V., Malik, H., Chandel, S.S., 2018. Daily array yield prediction
Dev. 47, 9–16. of grid-interactive photovoltaic plant using relief attribute evaluator based
Şenol, M., Abbasoğlu, S., Kükrerb, O., Babatunde, A.A., 2016. A guide in installing Radial Basis Function Neural Network. Renew. Sustain. Energy Rev. 81 (2),
large-scale PV power plant for self-consumption mechanism. Sol. Energy 132, 2115–2127.
518–537. Yao, W., Li, Z., Wang, Y., Jiang, F., Hu, L., 2014. Evaluation of global solar radiation
models for Shanghai, China. Energy Convers. Manage. 84, 597–612.

7314