Renewable Energy 193 (2022) 857e869

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

New model to study the outdoor degradation of thinefilm

photovoltaic modules
Michel Piliougine a, *, Paula Sa

nchez-Friera b, Giovanni Petrone a,

Sa
Francisco Jose
nchez-Pacheco c, Giovanni Spagnuolo a, Mariano Sidrach-de-Cardona d
a  degli Studi di Salerno, 84084, Fisciano, SA, Italy
Dipartimento di Ingegneria dell’Informazione ed Elettrica e Matematica Applicata, Universita
b
n, Spain
Solar PV Consultancy Services, Solkeys, 33204, Gijo
c
Departamento de Tecnología Electro
nica, Universidad de Ma
laga, 29071, Ma

laga, Spain
d
Departamento de Física Aplicada II, Universidad de Ma
laga, 29071, Ma

laga, Spain

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

Article history: The performance of four thin-film photovoltaic modules is analyzed after an initial stabilization period
Received 9 February 2022 and a subsequent outdoor exposition. The seasonal variations and the degradation rates of a single-
Received in revised form junction hydrogenated amorphous silicon (a-Si:H) module, a tandem amorphous microcrystalline Sili-
6 May 2022
con (a-Si/mc-Si) module, a heterostructure cadmium sulfide-cadmium telluride (CdS/CdTe) module and a
Accepted 11 May 2022
copper indium gallium selenide (CIGS) are examined and correlated to spectral changes. The IeV curves
Available online 19 May 2022
have been measured every 5 min; the electrical parameters and parasitic resistances have been iden-
tified. By exploiting a number of experimental measurements acquired within a narrow interval of
Keywords:
Amorphous silicon
irradiance and cell temperature, a novel mathematical model has been developed and fitted: it considers
Cadmium telluride a stationary seasonal variation component and a linear long-term degradation component. The results
Copper indium gallium selenide show annual power degradation rates of 4.0% for the a-Si:H module, 3.4% for the a-Si/mc-Si and 3.5% for
Microcrystalline silicon the CdS/CdTe, whereas for the CIGS module the annual degradation is not significant, i.e. 0.2%.
Photovoltaic degradation © 2022 Elsevier Ltd. All rights reserved.
Thin film

1. Introduction The degradation mechanisms associated to thinefilm modules

have been the subject of many works for the last years, e.g.
Thinefilm PV modules had an important presence in the global Refs. [4e26], although less than those ones affording the degra-
market a decade ago, reaching about 17% of the global production in dation of ceSi modules [27]. Thinefilm PV module degradation
2009 [1] and then gradually loosing share due mostly to the im- rates are in general higher, and with wider dispersion than their
provements and cost reduction of the crystalline silicon (ceSi) crystalline counterparts [28], this being one of the factors that have
technologies. Nevertheless, they still constitute today a viable op- hampered the market growth of these modules. However, some
tion, which is particularly attractive in certain applications with a reports [29,30] have stated the that thinefilm technologies may
large growth potential such as building integrated photovoltaics have better stability. In any case, it is highly desirable to conduct
[2,3]. The market share of thinefilm technologies in the global PV further studies to gain a deeper insight into the degradation of
market is expected to be limited. Nevertheless, understanding the thinefilm modules. This will contribute to improve their reliability
degradation mechanisms of a-Si and micromorph modules could and reduce uncertainties in the estimation of their longeterm en-
provide insights on the behaviour of high efficient silicon ergy production in real systems.
heteroejunction technologies, which also contain a-Si layers, and Table 1 summarizes some thinefilm module degradation rates.
are recently attracting a lot of interest. Even for the same technology, there is a significant dispersion of
the levels of degradation. Considering aeSi:H, many papers
report annual degradation rates around 2% [12,14,18,19] or even
lower [6,16,20], but other studies report rates ranging between 5
* Corresponding author.
E-mail addresses: [email protected] (M. Piliougine), paula@solkeys. and 7% [4,17]. Lower degradation rates are obtained for modules
com (P. S
anchez-Friera), [email protected] (G. Petrone), [email protected] based on tandem cells of amorphous and microcrystalline silicon

nchez-Pacheco), [email protected] (G. Spagnuolo), [email protected]
(F.J. Sa (aeSi/mceSi), lying in the range from 1.5 to 2.2% [8e11,13],
(M. Sidrach-de-Cardona).

https://doi.org/10.1016/j.renene.2022.05.063
0960-1481/© 2022 Elsevier Ltd. All rights reserved.

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

Table 1
Summary of different studies of degradation for thinefilm modules.

Date Reference Technology Brand/model/identification Years Loadb
%Pmax/year

2021 Cassini et al. [4] aeSi:H Module 3a 15 inverter 5.5

CdTe Module 4a 7 inverter 5.6
2021 Kumar et al. [5] CdTe Specimen CdTe-6a 1 ND 1.37
2020 Singh et al. [6] aeSi:H same brand/model, a-Si_1a 3 ND 1.32
same brand/model, a-Si_2a 3 ND 1.12
same brand/model, a-Si_3a 3 ND 1.04
2018 Kichou et al. [7] CdTe Calyxo/CX3-80 2 MPP 5.55
2018 Aarich et al. [8] aeSi/mceSi Array series/parallela þ2.5 inverter 1.5
2017 Limmanee et al. [9] aeSi/mceSi Thin Si 1a 4 inverter 2.1
aeSi/mceSi Thin Si 2a 4 inverter 1.8
CI(G)S CIGSa 4 inverter 1.7
2017 Ozden et al. [10] aeSi/mceSi Kaneka/U-EA105 4 inverter 1.88
CdTe Abound/AB1-67B 4 inverter 10.6
2017 Tahri et al. [11] aeSi/mceSi Sharp/NA-121 3 ND 1.73
CI(G)S Solar Frontier/SF150eS 3 ND 2.34
2016 Kichou et al. [12] aeSi:H Array series/parallela 3.5 inverter 2.30
2016 Kichou et al. [13] aeSi/mceSi Array series/parallela 3.5 inverter 2.20
2016 Silvestre et al. [14] aeSi:H Kaneka/G-EA60 5 ND 2.28
aeSi/mceSi Sharp/NA-121 5 ND 2.72
CdTe Fist Solar FS-270 5 ND 4.55
CI(G)S Shell Powermax Ultra80C 5 ND 1.04
2015 Ferrada et al. [15] aeSi/mceSi Array series/parallela z1.5 inverter 2.7 ~3.9
2015 Kyprianou et al. [16] aeSi:H Mitsubishi/MA100T2 8 inverter 1.38
CdTe Fist Solar FS-60 8 inverter 1.96
CI(G)S Würth Solar/WS11007-75 8 inverter 2.33
2014 Sharma et al. [17] aeSi:H Array series/parallela þ2 inverter 6.4
2014 Ye et al. [18] aeSi:H Specimen 6a 3 MPP z2
CI(G)S Specimen 9a 3 MPP z6
CdTe Specimen 10a 3 MPP z2
2014 Makrides et al. [19] aeSi:H Mitsubishi/MA100T2 5 inverter 1.87
CdTe First Solar/FS-60 5 inverter 2.27
CI(G)S Würth Solar/WS11007-75 5 inverter 1.95
2013 Gottschalg et al. [20] aeSi:H Specimen aeSie1a þ4 ND 1.26
CI(G)S Specimen CIGSe1a 5.5 ND 2.74
CI(G)S Specimen CIGSe2a 5.5 ND 0.42
2012 Strevel et al. [21] CdTe Solar Cells Inc. (act. First Solar) 17 inverter 0.53
2011 Jordan and Kurtz [22] CI(G)S ShellSolar/PowerMax Eclipse 80-C þ5 inverter 0.32
2008 Del Cueto et al. [23] CI(G)S Specimens 1988:Aa þ17 resistor 0.90
CI(G)S Specimens 1990:Aa 16.5 resistor 0.21,0.32
CI(G)S Specimens 1992:Aa 15.5 resistor 0.43
CI(G)S Specimens 1994:Aa 13 resistor 0.95,1.08
CI(G)S Specimens 1997:Aa 10 resistor 1.27,1.35,2.10
CI(G)S Specimens 1997:Aa 10 resistor 2.10
CI(G)S Specimens 1998:Aa þ9 resistor 2.25
CI(G)S Specimens 1997:Aa 6 resistor 1.27,1.35
CI(G)S Specimens 2002:Ba 5.5 resistor 0.89,1.80
CI(G)S Specimens 2002:Ba þ4 resistor 0.65
CI(G)S Specimens 1998:Aa 4 resistor 2.14
CI(G)S Specimens 2004:Aa 3 resistor 2.6,2.9,3.8,4.7
2006 Durisch et al. [24] CI(G)S Siemens/ST40 1 ND 8.4,9.5
2005 Adelstein and Sekulic [25] aeSi:H Array series/parallela 1 inverter 7.3
(same installation, following years)a 5 inverter 1.7
2001 Marion et al. [26] CdTe Solar Cells Inc. (act. First Solar) 4 inverter 0.6
a
The actual brand/model are not defined in the referenced paper.
b
It specifies if the modules are normally (when they are not being measured): ”resistor” (with a fixed resistor), ”MPP” (at its maximum power point with an optimizer
circuit), ”inverter” (several of them connected as an array to a grid inverter), or ”ND” (not defined in the referenced paper).

although higher rates, between 2.7 and 3.9%, are reported in modules variables [32]. In round robin measurements conducted
Ref. [15]. by different test laboratories, divergences between 3 and 6% were
The dispersion of the degradation rates for Cadmium Telluride found for thinefilm PV modules, whereas for crystalline silicon
(CdTe) is even higher. Whereas some authors [21,26] show very low modules, divergences were under 1% for the same laboratories. In
degradation rates, i.e around 0.6%/year, some works evidence rates general, it is known that the output of a thinefilm module depends
between 1.4% and 10% [4,5,7,10,11,14,16,18,19]. Regarding CI(G)S on its actual operating conditions, but it is also affected by its past
technologies, a wide difference in the results documented by history, both in terms of temperature and irradiance [33], that may
different authors is evident [9,11,14,16,20,22,31], with annual affect their performance permanently or reversibly.
degradation rates ranging from 0.2 to 2.6%. In general, significantly The initial stabilization period of the modules is normally
lower rates are reported in comparison to the other thinefilm observed in all thinefilm PV technologies. As for the aeSi based
technologies with a few exceptions [18,24]. The large dispersion technologies, the welleknown Staebler-Wronski effect [34] leads to
found in reported degradation rates is partly due to the difficulties about 20% of the initial efficiency drop during the first weeks of
associated to ensure a reliable measurement of thinefilm PV exposure to sunlight due to the formation of metaestable defects in

858

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

the aeSi:H semiconductor. The phenomenon depends on the level thinefilm degradation at outdoor conditions should consider
of irradiation and also on the device temperature [35]. In fact, the measurements performed under a similar spectrum, e.g., acquired
initial performance can be recovered by annealing above certain during the same month of different years.
value of temperature, i.e. from z 130  C. For lower temperature In this work, the results achieved by applying a rigorous pro-
levels there is a partial recovery depending on the thermal history cedure to analyze the evolution of the outdoor performance of four
of the device and on the climate conditions. The stabilization period different types of thinefilm modules are presented. The modules
for aeSi is typically considered to be of about 6 weeks [36e39], are of the following technologies: single-junction hydrogenated
during which the most significant power decay takes place. Other amorphous silicon (aeSi:H), tandem of amorphous and micro-
studies report longer periods for a complete stabilization crystalline silicon (aeSi/mceSi), heterostructure of cadmium sulfide
[13,40,41]. Other thinefilm modules suffer a strong initial degra- and cadmium telluride (CdS/CdTe) and copper indium gallium
dation due to light exposure. For CdTe, 6-week stabilization period selenide (CIGS). The modules under study were installed on the
is commonly accepted. Del Cueto et al. [42] state an initial drop
laga in the
roof of the Photovoltaic Laboratory of the University of Ma
after the first 400e600 h of sun exposure, which is equivalent to South of Spain and they have been monitored during a period of
6e8 weeks. Deline et al. [43] recommend an initial period of around 20 months, after a stabilization period of 8 weeks. The re-
120 KWh/m2, which is equivalent to 6 weeks assuming a minimum ported degradation rates quantify the evolution of these indicators:
daily irradiation of 2.78 kWh/m2, estimated for our site (Ma
laga, the maximum power (Pmax), the shortecircuit current (ISC), the
Spain) in the worstecase month [44]. Regarding CIGS, there is not openecircuit voltage (VOC), the current at the MPP (Imax), the
agreement in the literature [45]. Some works show a stable per- voltage at the MPP (Vmax), the fill factor (FF), the series resistance
formance since the first use [14]; others conclude that at least 1 (RS), and the shunt resistance (RSh), these last two referred to the
year is necessary to reach a stable behavior [46]. single diode model (SDM) [51].
Despite its importance, not all the studies state if the initial Aimed at distinguishing between the seasonal annual variations
stabilization period is included or not in the reported degradation and the longeterm degradation of the previous indicators, a novel
rates. For example, in Demirtaş et al. [47] the degradation rate is methodology is proposed, based on fitting the measured data to a
estimated during the first year of operation for two CdTe and CIS nonelinear model that can deal with noneequispaced data and
arrays, by obtaining 22% and 32% respectively. These very high with missing information. This model assumes that the amplitude
values suggest that the initial drop is included in the degradation of the seasonal trend is a percentage of a base value, which in turn
rate, which would not be representative of a longeterm study. degrades over time. The fitting parameters obtained will be used to
Another frequent issue in thinefilm PV degradation analyzes is estimate the annual degradation rates and correlate the behavior of
related to the accuracy of the measurements. It is not uncommon to the modules to the spectral variations.
find studies [48] based on measurements performed by IeV The paper is organized as follows: In Section 2, the methodology
portable curve tracers, which are designed to characterize whole employed for the measurement and parameters extraction is
strings. Unfortunately, they offer an insufficient accuracy to specify described. The results of the evolution of the electrical parameters
the level of degradation of an individual module, especially if the and parasitic resistances are presented in Section 3. Sections from
exposure period is short: the actual degradation is often smaller 3.1 to 3.4 analyze the obtained results for each technology, con-
than the measurement error. Moreover, as many of these devices trasting them with previous literature. Finally, Section 4 summa-
are based on a capacitive load, they are not fast enough to accu- rizes the main conclusions of this work.
rately sample the IeV curve in proximity of the shortecircuit cur-
rent, thus further reducing the accuracy of the ISC and RSh 2. Methodology
estimations. A too fast acquisition of the IeV curve, as it happens
with some solar simulators, may lead to transient effects (e.g. the 2.1. Description of PV modules
direction of the voltage sweep changes the shape of the curve),
with consequent errors in the estimation of Pmax, in particular One specimen for each technology has been used, labelled as
when measurements concern high capacitance thinefilm tech- M1, M2, M3, and M4. The main specifications of the modules under
nologies [49]. study are summarized in Table 2. M1 and M2 are based on amor-
Outdoor measurements are usually translated to Standard Test phous silicon, one of the most sustainable PV technologies, due to
Conditions (STC) by using correction procedures. The latter ones its abundance, low temperature coefficients, small amount of raw
introduce additional uncertainties in the final values, which can be material required and loweenergy manufacturing process [52].
sometimes larger than the determined degradation rates. For Amorphous silicon is usually hydrogenated, as the incorporation of
example, the procedures 1, 2 and 4, described in the IEC atomic hydrogen passivates the dangling bonds typical of the
60891:2021 [50], as most of other methods in the literature, amorphous structure thus improving the cell efficiency. The hy-
require, in addition of the temperature coefficients, other internal drogenated amorphous silicon (aeSi:H) material has a high direct
parameters that are hard to be obtained and whose values are band gap of 1.7 eV which leads to high absorption in the short
usually affected by a significant approximation. The introduced wavelengths. M1 is a singleejunction aeSi:H module by Kaneka
uncertainty can be especially large in climates where it is not with a p-i-n cell configuration and a nameplate stabilized efficiency
possible to reach high irradiance values, e.g. close to 1 000 W/m2, of 6.3% [53]. M2 is a tandem a-Si/mc-Si module by Phoenix Solar
with low cell temperatures, so that the temperature gap to correct which includes a top aeSi cell and a bottom mceSi cell, and has a
(e.g. from 50  C to 25  C) is very large. efficiency of 8.5% [54]. Both M1 and M2 have power tolerances
As a further element affecting degradation analyzes based on of þ10%/
5% and power warranties of 80% at 25 years. They were
outdoor measurements, variations of the spectrum have to be built with a glassefoil superstrate configuration. The Kaneka
accounted for. Some thinefilm technologies show a narrow spectral module includes a 5 mm float glass whereas the Phoenix Solar
response, which leads to significant performance changes due to module is built over a 4 mm loweiron front glass. Both have EVA as
spectral variations throughout the year. For example, in King et al. encapsulant, a fluorineebased foil backsheet and aluminum
[41] it is stated that the energy production of an aeSi module could frames.
increase even by a 2% in summer with respect to its output under M3 is a CdS/CdTe module by First Solar with a nameplate effi-
AM1.5, with a decrease of a 5% in winter. As a result, a valid study of ciency of 10.07% and a power tolerance according to the
859

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

Table 2
Specifications of the photovoltaic modules under study.

Module M1:a-Si M2:a-Si/mc-Si M3:CdTe/CdS M4:CIGS

Manufacturer Kaneka [53] PhoenixSolar [54] FirstSolar [55] Solyndra[56, 57]

Model GEAe060 PHXe120eLV FSe272 SL-001-157C
Number of cells 108 96 116 150
Efficiency [%] 6.3 8.51 10.07 7.99
Temperature coefficient a [%/ C] þ0.075 þ0.07 þ0.04
0.06
Temperature coefficient b [%/ C]
0.305
0.30
0.25
0.38
Temperature coefficient g [%/ C]
0.23
0.24
0.25
0.35
NOCT [ C] e 44 45 45
Electrical Parametersa STC NOCT STC NOCT STC NOCT STC NOCT
Shortecircuit current ISC [A] 1.19 e 3.34 2.70 1.23 1.01 2.54 1.95
Openecircuit voltage VOC [V] 92 e 59.20 55.10 88.7 82.5 97.60 88.30
Maximum power Pmax [W] 60 e 121.0 91.80 72.5 54.4 157.0 113.4
Current at maximum power Imax [A] 0.90 e 2.69 2.19 1.09 0.87 2.20 1.69
Voltage at maximum power Vmax [V] 67 e 45.00 41.90 66.6 62.4 72.30 67.10
a
STC: G ¼ 1 000 W/m2, Tcell ¼ 25  C, AM1.5 d NOCT: G ¼ 800 W/m2, Tair ¼ 20  C, Windspeed ¼ 1 m/s, AM1.5.

manufacturer of ±5%. It is a doubleeglass frameless module, with The output signals of all the sensors are connected to a Compact
glass panes of 3.2 mm thickness, heat strengthened at the front and FieldPoint™ cFPe2120 controller and acquisition system from
tempered at the back, laminated with an encapsulant and an edge National Instruments [63]. The system is connected to the local
seal [55]. M4 is a CIGS module, material that has a very high ab- Ethernet in such a way any computer can retrieve the measure-
sorption coefficient and also a large direct band gap. Manufactured ments in real time. The IeV curve acquisition system was described
by Solyndra, with a total area nameplate efficiency of 7.99% and a in a previous paper [64]. A fourequadrant power supply performs
power tolerance of ±4% [56]. This module has a special construction the voltage sweep at the PV module terminals in a suitable voltage
which was conceived to improve generation at low incidence an- range. The IeV samples are acquired by a pair of multimeters
gles. It is built from a series of tubes, which are composed of two triggered by a common square signal to ensure the simultaneity of
cylinders one inside the other. The CIGS cell is deposited over the both readings. The acquired IeV curves are stored in a relational
internal cylinder and optically coupled to the outer cylinder. The database and can be downloaded and exported for further treat-
tubes are hermetically sealed with metal caps at the ends of the ment through a web application [65]. The IeV acquisition time and
tubes. the number of points can be configured, setting these parameters to
1 s and 100 points respectively. In order to share the equipment
2.2. Experimental seteup between all the installed PV modules, a relay box is used to select
automatically the specimen to be measured, in such a way all of
The modules were mounted on a fixed openerack structure them are measured in sequence every 5 min. The PV modules have
facing to south with an inclination of 21. The installation is in remained at openecircuit condition when they were not measured.
Ma
laga (latitude 36.7 N, longitude 4.5 W, altitude 50 m) in the A detailed analysis of the uncertainties of this measurement
south of Spain, very close to the Mediterranean Sea and with typical system, used in Ma
laga for the last years, is provided in Ref. [66].
meteorological conditions of 1890 kWh/m2 of annual horizontal The shortecircuit current ISC is estimated with an expanded un-
global irradiation, average daytime ambient temperature of 18  C, certainty of 0.8% whereas the expanded uncertainty for the
relative humidity of 63% and wind speed of 1 m/s. openecircuit voltage VOC is 0.5%. However, the estimations of Pmax,
The meteorological variables were acquired by a weather station Vmax and Pmax have expanded uncertainties of 1.1% (coverage factor
composed of an anemometer and a wind vane (Young 03002L), a k ¼ 2 in all the cases accordingly to ISO/IEC Guide 98-3 [67]). Finally,
combined sensor of air temperature and relative humidity (Young the expanded uncertainty for the calculated values of FF, RS and RSh
41382LC) and two irradiance sensors on the module plane: a are 1.6%, 3% and 7% respectively.
thermoeelectric pyranometer Kipp & Zonen CMe21 [58] and a
reference cell from the manufacturer Atersa [59]. Whereas the 2.3. Measuring and data extraction procedure
former provides the irradiance used as reference value, the latter
fasteresponse sensor is only used to discard those IeV curves Before the start of the measuring campaign, the modules were
measured under changing conditions of irradiance. stabilized by leaving them outdoors for an initial period of 8 weeks.
The cell temperature is monitored with the use of Pt100 Resis- During this stabilization period the PV modules have been also at
tance Temperature Detectors (RTD) using a fourewire connection openecircuit condition without any load connected to them. As it
to avoid ohmic losses. These sensors are encapsulated by a silicone has been stated in Section 1, the purpose of this period is to avoid
sheath allowing a better coupling to the back surface of the PV the strong power drop during the first weeks of exposure, assuming
modules [60]. Several sensors are used for each module and an that the complete stabilization period could take much more time.
average cell temperature is calculated, following the schema After stabilization, the IeV curves of the modules have been
depicted in IEC 60891:2021 [50], to take into account edge effects. measured at 5-min intervals during the sun hours for the whole
The spectrum of the solar irradiance is also monitored using a period under assessment, that lasts around 20 months. Simulta-
spectroradiometer EKO MSe710 installed on the module plane, neously with the IeV curve, the following magnitudes were
covering from 350 to 1100 nm. Although this range could not be measured: irradiance on the module plane, ambient temperature,
enough for all the technologies under study (e.g. aeSi: cell temperature (at several points), speed and direction of the
300e790 nm or CIGS: 370e1240 nm [61]), the solar spectrum for wind, relative humidity and solar irradiance spectrum. All the
the portions not covered is almost insignificant for our purposes modules and the sensors were regularly cleaned to avoid soiling.
(the irradiance between 300e350 nm and 1100e1240 nm are only The estimation of the electrical parameters from the discrete
1.3% and 4.9% of the total solar irradiance [62] respectively). points of the IeV curves is done following the indications given by
860

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

Emery [68]. Once the main electrical parameters have been esti- literature that are aimed at separating the seasonality due to the
mated, a parameter identification method based on the thermal and radiative history of the module from those related to
singleediode model can be applied to estimate the values of the the variability of the actual incident spectrum. Pierro et al. [79]
series and shunt resistances, such as the one described by Toledo propose a model to estimate the individual contribution on the
and Blanes [69]. This analytical method requires the coordinates output performance of each influential variable.
and the derivatives of four points of the IeV curve; therefore, ISC, In order to distinguish between the seasonal annual trend and
VOC, Pmax, and an additional point, selected by means of the the longeterm underlying degradation of thinefilm devices, the
a
power function described in the same paper, have been used. A sequence of values of each electrical parameter should be fitted to a
value of a ¼ 10, thus giving a point between the MPP and the model that allows to separate each component. Some approaches
openecircuit point, has been settled. are framed inside the timeeseries forecasting methods, with very
In addition, from the measurements of the solar spectrum, the low computational complexity, that have previously been used to
associated average photon energy (APE) values were computed. The study PV degradation, such as Autorregressive Integrated Moving
APE is defined as the mean energy value (expressed in eV) of the Average (ARIMA) [80] or SeasonaleTrend decomposition by Loess
photons from a given spectral distribution [70] and it is used to be (STL) [81]. These methods require daily or monthly time series, in
estimated referred to a finite wavelength interval [la, lb] as: such a way, after performing the filter by irradiance and tempera-
ture, it is necessary to integrate the instantaneous measurements to
ð lb
obtain a discrete equispaced sequence of values, introducing
Gl ðlÞdl
l additional uncertainty.
APEllb ¼ ð la (1) Another approach that can deal with noneequispaced data and
a b
Gl ðlÞ
dl with missing information (e.g., several days without measure-
la εl
ments) was proposed by Ref. [82], consisting on a nonelinear
model which include several additive terms, including a sinusoi-
where Gl is the spectral irradiance on the module plane for a spe-
dal one to catch the seasonal trend. A fitting routine for nonelinear
cific wavelength l (in Wm
2/nm) and εl is the energy (in eV) of a
models is available within any mathematical software, e.g., lsqcur-
photon of wavelength equal to l. This APE indicator is a useful
vefit in Matlab R2021b [83], that implements the trust-
parameter that has been widely used to represent as a unique scalar
eregionereflective algorithm to set a value for each free parameter
the spectral irradiance function [71,72]. In our case, and due to the
of the model within a feasible interval. Despite its high computa-
limitations of our spectroradiometer, la and lb have been set to 350
tional requirements, with the computing power of today's micro-
and 1100 nm respectively.
processors, this task requires only a few seconds of execution.
Ideally, for estimating the degradation rates, the operating
In the same line of that previous work, we propose the following
conditions of the module should be the most stable and referred to
new model, in which the variation of each indicator F(t) can be
the same values throughout the period. In addition, to compare the
modeled through (eq. (2)), assuming that t is expressed in years and
obtained rates with previous works, those measurements should
t0 is the date of the first measurement:
be related to STC: GSTC ¼ 1 000 W/m2 and TSTC ¼ 25  C. Unfortu-

laga makes such a
nately, the climate of the installation site in Ma
combination of G and T practically impossible. By choosing the
experimental measurements referring to GSTC implies to neglect FðtÞ ¼ ðK0 þ K1 ,ðt
t0 ÞÞ, ð1 þ K2 ,sinð2p,ðt
t0 þ K3 ÞÞ
most of the measurements taken in winter period. Therefore, it has |fflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
long
term seasonal
term
been decided to select, within the huge number of experimental
measurements available, those ones referring to an irradiance level (2)
close to GNOCT ¼ 800 W/m2, thus ensuring that data referring all the
where the first term represents a linear longeterm degradation,
seasons of the year are used. After having creamed off the data
and the second term contains the information on the seasonal
according to this filter, the center of the temperature range has
variations, which are mostly due to spectral effects or annealing.
been fixed individually for each different PV module, by taking the
There are some important differences between the proposed
mean temperature value Tc among all the samples within the
model and the one used by Ref. [82]. On the one hand, the model
irradiance range. Regarding the radius of the intervals of G and T,
proposed in that work assumes a logarithmic longeterm compo-
30 W/m2 and 1.5  C respectively have been considered, which are
nent in order to deal with the initial strong degradation. However,
exactly the expanded uncertainty that are estimated for the read-
as the first 8 weeks of exposure are not included in the present
ings of the pyranometer (Kipp & Zonen CMe21) and the cell tem-
study, the model presented in Ref. [82] has been simplified by
perature sensor (RTD Pt100 class A).
approximating the longeterm component by a linear trend. On the
other hand, it has physical sense that the amplitude of the seasonal
2.4. Degradation model trend, which is the factor multiplying the sinusoidal term, is a
percentage of a base value determined by the longeterm compo-
In some previous studies, the degradation along the time is nent. Therefore, instead of using an additive model of the
considered as linear at a first approximation, so that a Simple Linear longeterm and the seasonaleterm, in (eq. (2)) it is proposed a base
Regression (SLR) could be used [73]. However, PV modules show value that could change over time, K0 þ K1 , (t
t0), and let the
different seasonal patterns, especially thinefilm devices. This seasonal part vary as a ratio K2 of that base value. In this way,
behavior has been known since the 1990s, specially for aeSi:H, as it assuming that t is expressed in years, the slope K1 represents the
is stated by Rüther and Livingstone [74], and other previous works absolute variation of the parameter for each year (e.g., in case of
[75e77]. These are due to variations in the solar spectrum along the Pmax, its unit is W/year), whereas K0 is equal to the value of such
year and, for amorphous silicon modules, to thermal annealing, parameter at the initial time t0. Therefore, a relative annual
which can partially recover the initial losses due to light induced degradation rate can be estimated as a ratio between K1 and K0, i.e.,
degradation [78]. DR(%/year) ¼ 100 , (K1/K0). Finally, the parameter K3, which is
The history profile of a PV thinefilm module and its impact expressed in years, represents a phase value that allows a delay
depend on the location. However, there are some attempts in the between the behavior of the different electrical parameters.
861

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

for every PV module, a different center of the temperature interval

has been set, assuming the mean temperature value Tc among all
the samples within the previous irradiance range (with a fixed
radius of 1.5  C for all the cases). The results shown in Fig. 1 and in
Figs. 2e5 and indicate a strong correlation of the modules electrical
parameters and parasitic resistances with the month of the year
and, consequently, with the APE. As it has been stated in Section 1,
this seasonal variation makes impossible to compare two mea-
surements taken on different days of the year to quantify the
degradation. In general, the electrical parameters experiment a
longeterm degradation that is superposed to the seasonal trend.
The experimental results of the main electrical parameters have
Fig. 1. Evolution of the APE during the exposure period, restricted to G ¼ 800 ± 30 W/ also been fitted to the degradation model described in Section 2.4.
m2
Tc ¼ 38 ± 7  C. Tables 4e7 show, in addition to the Ki parameters estimated in each
fitting, the value of the degradation rate calculated for each elec-
trical parameter and each module type.
3. Results and discussion
3.1. Amorphous-Si module
First, the evolution of APE value is shown in Fig. 1 restricted to
the ranges G ¼ 800 ± 30 W/m2 and Tc ¼ 38 ± 7  C, that covers the
For the module M1, of a-Si:H technology, all the electrical pa-
irradiance and temperature ranges assumed for all PV modules
rameters (ISC, Imax, VOC, Vmax, Pmax) show seasonal variations which
under study. During the summer months the APE is higher, which
are directly correlated to the APE variation, i.e. they increase in the
indicates a solar spectrum shifted to the blue, whereas in the winter
summer months and decrease in winter. The value K2(Pmax) (see
the APE is lower, indicating a shift towards the red. To have a
Table 4) gives as a measurement of the amplitude of this seasonal
reference of the evolution of the APE, these data have been fitted to
variation, leading to a 6.4%, which have been attributed to different
the model in Section 2.4. However, as the solar spectrum is not
effects such as temperature annealing [84] and spectral effects [85].
assumed to be affected by any longeterm variation, in this case we
Rüther et al. [86] suggest that one of the reasons for this behavior
propose to fit only the seasonal part of the model, in other words,
could also be a field driven transport in the amorphous material.
before performing the fitting procedure we will set the value
This seasonal change is so strong that mask the longeterm degra-
K1(APE) ¼ 0. The parameters Ki in Table 3 describe the behavior of
dation, but the model is able to separate them.
the APE during the exposure period. The value K0(APE) ¼ 1.893 is an
We must highlight the fact that the value of K3(ISC) and the same
estimation of the APE (in eV) on t ¼ t0, i.e., on the first registered
parameter obtained when fitting the APE values are very close. This
day. The value K2(APE) ¼
1.07e-2 is the amplitude of the seasonal
means that the current output of this type of modules has a little
variation of the APE. Finally, K3(APE) ¼ 0.175 (expressed in years) is
phaseeshift with respect the variations of the solar spectral dis-
the delay of the initial day t0 with respect the natural beginning of
tribution. The decrease of the shortecircuit current of aeSi mod-
the APE sinusoidal oscillation (around 64 days).
ules when the spectrum shifts towards longer wavelengths (red)
For comparative purposes, in this paper the APE evolution
has been reported before in other studies [84] and it is a direct
(calculated on the complete wavelength range provided by the
consequence of the lower spectral response of these modules in the
spectrometer) is provided to explain the annual seasonal trend of
longer wavelength range, which is caused by their large direct band
the main electrical parameters. It could be possible to cancel the
gap.
seasonal spectral influence by using a correction based on APE data,
In our case, the annual degradation rate is 4.0%, that is within
but in that case it would be necessary to apply different wavelength
the upper range stated by other works in the literature referred to
limits depending on each technology under test. However, in this
this technology. First, some studies have analyzed the dependence
work the APE data have not been used to perform any correction to
of the degradation rate on the electrical load, showing that
the measurements to avoid additional uncertainties on the degra-
degradation for aeSi modules can be larger in openecircuit con-
dation rates.
ditions [87], as it is our case. In the review work conducted by
The evolution of the main electrical parameters and the parasitic
Jordan and Kurtz [88], the reported degradation rates for aeSi PV
resistances for each of the four modules, extracted from the
modules were highly dispersed and going up to 4.5%. Returning to
experimental measurements, is shown in Figs. 2e5. The dotted
Table 1, it is possible to see lower estimations [6,12,18e20], but also
lines show the interval in which the majority (indicate a quanti-
higher ones [4,17,25].
tative portion) of the points are contained and they are represented
Silvestre et al. [14] study a specimen of the same manufacturer
to guide the eye.
and model, obtaining a lower annual degradation rate (around
Before fitting the degradation model of Section 2.4, a first irra-
2.3%). This module M1 has obtained the worst power degradation
diance filter has been applied to the dataset of measurements of all
rate of all the modules under study, due to a combination of a high
PV modules (G ¼ 800 ± 30 W/m2). However, a second filter based
decrement of ISC (1.3%/year, more than any other module), a
on the cell temperature has also been applied to each PV module. In
decrement of VOC (1.0%, again more than any other module), and an
order to ensure a high number of selected points to fit the model,
increase of the RS (8.6%/year). From Fig. 2g, it can be seen that RS
maintains stable during the cool months, whereas the increments
happen during the hottest months of the year. This effect due to
Table 3 temperature is inverse to the effect of the annealing, so the bene-
Fitting parameters Ki for data of APE.
ficial effects of temperature are counteracted in this way. On the
K0 K1 K2 K3 other hand, the RSh has not experienced a significant variation (only
APE 1.893 eV e
1.07e-2 þ0.175 years a decrement of 2.7%, that is very low assuming the high uncertainty
associated to the estimation of this parameter).

862

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

Fig. 2. Evolution of main electrical parameters and parasitic resistances for module M1 (aeSi:H).

3.2. Tandem AmorphouseSi and MicrocrystallineeSi module aeSi modules, in this case only the currents ISC, Imax are directly
correlated with APE, whereas the voltages VOC, Vmax show the
In case of the module M2 (aeSi/mceSi), again the studied elec- opposite tendency, having maximum values when APE is lowest.
trical parameters are affected by a strong seasonal trend, as it can be The amplitude of this seasonal trend in terms of power is given by
seen in Fig. 3 and Table 5. However, whereas the maximum values the value K2(Pmax) ¼ 1.90%, much lower than for the other studied
of ISC and Imax are in summer and the minimum ones in winter, the technologies. This opposite tendency of ISC and VOC in CdTe modules
behavior VOC and Vmax leads to the highest values in the autumn has also been observed in previous literature [89]. Instead of
and the lowest values in spring. The magnitude of the seasonal reporting the evolution of the values of Pmax, in that paper those
oscillations in terms of maximum power is quantified by K2(Pmax) data are expressed in terms of instant efficiency (the same infor-
(3.1%). Part of it is due to the spectral variation, and the rest should mation normalizing by the module area).
be due to annealing. As it can be seen, the influence of the The obtained annual degradation rates for the module M3 is
annealing is less than in case of aeSi:H, as it is stated by Silvestre shown in Table 6. For the maximum power, we obtained an annual
et al. [14]. The annual degradation in terms of ISC is almost as high as degradation rate of 3.5%/year for the CdS/CdTe module. As ISC has
in case of aeSi (1.0%/year), but the degradation rate of VOC is sig- experience a very low degradation rate (only 0.4%/year) and VOC
nificant lower (only 0.3%). However, the most serious problem of does not have a strong decrement (0.7%/year), the main contribu-
this PV module was the very high increment of the series resistance tion to the power drop is related to the evolution of both parasitic
RS (23%/year), much more than any other module under study. On resistances: an important increment of the RS (13%/year), and
the other hand, RSh has experienced a moderate decrement (7.5%/ especially, a very high degradation of the RSh (26%/year). This
year). In balance, the power annual degradation rate of this M2 abnormal increment of the shunt resistance could be due to the fact
specimen is 3.4%, significantly higher than the results stated by that this module is not provided with a frame, in such a way it is
Refs. [8e11], but in the same range than the work by Ref. [15]. more prone to have problems related to humidity that significantly
However, some surveys [29,30] report even negative degradation influences this parameter. As it happens with module M1 and M2,
rates, although it could be due to undererating, lightesoaking or the increments of RS take place in the summer and early autumn,
spectral mismatch issues. whereas in the rest of the year the value is very stable.
Although there are suspicious papers [21,26,90] giving power
3.3. CdTe/CdS module drops around 0.6%/year (some authors have a contractual rela-
tionship with the manufacturer), many works report annual
Fig. 4 shows the tendency observed for the electrical parameters degradation rates for Pmax from 2 to 6% [4,7,14,18,19], that is in
of the module M3, based on CdTe. As opposed to what happens in agreement to our result.

863

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

Fig. 3. Evolution of main electrical parameters and parasitic resistances for module M2 (aeSi/mceSi).

3.4. CIGS module from Table 7, the magnitude of K2(Pmax) ¼ 6.67% indicates a strong
seasonal variability that cannot be explained by the APE variation.
Some previous considerations should be taken into account. This Therefore, there must be other effects affecting the behavior of the
cylindrical module has a construction not comparable to the one of the CIGS module.
conventional PV modules having very different mechanical, electrical, On the one hand, the series resistance of the CIGS module is
thermal and optical properties [91]. On the one hand, the irradiance decreasing in time. There are very few previous studies that analyze
received by this module cannot be estimated from the reading pro- the evolution of series and parallel resistances as a CIGS module
vided by the pyranometer installed on the same plane, because this degrades. For example, Del Cueto et al. [31] compare the trend of a
module receives light in a different way that the other planar modules couple of CIGS modules from two manufacturers exposed outdoor
and it is thought to be installed horizontally on the floor, not on an throughout 5.5 years, obtaining different results: whereas one of
openerack structure. On the other hand, due to its special shape, the them experiment an increment from 1.3 to 1.7 U (5.6% each year),
RTD sensors could not be installed in the same way that in the other the other one remains around 1.3 U all the period. On the other
specimens, and for that reason the value of cell temperature acquired hand, the behavior of RSh reflects an almost insignificant decrement
by the measurement system could not be comparable to the readings of this parameter (2.5%), taking into account that this is the one
of the other modules under study (in fact, for this M4 module the with highest uncertainty.
registered temperature is abnormally lower than the ones associated Degradation mechanisms in CIGS cells have been associated to
to M1, M2 and M3 under the same conditions). All these elements can shunting due to the formation of leakage paths across the p-n
explain the discrepancy between the measured values in our facilities junction resulting from defect states associated to lattice mismatch
and the specifications in Table 2. Even so, since these conditions have at the heteroejunction [84]. The manufacturer claims that, based
been maintained throughout the entire exposure period, it is possible on its own accelerated and onefield tests, the annual power
to determine a relative degradation rate. degradation rate is less than 0.5% [92]. This could be in line with the
Fig. 5 shows the tendency observed for the electrical parameters results presented in this paper, because the estimated degradation
of module M4 (CIGS). Like in the case of CdTe, ISC and Imax are is not significant (an annual degradation in power around 0.2%/
directly correlated with APE, whereas VOC and Vmax show the year). The degradation rates in term of current and voltage are also
opposite tendency, having maximum values when APE is lowest. very low (0.3%/year and 0.1%/year respectively). This also agrees
CI(G)S PV modules normally have a lower dependence on spectrum with the results discussed in Musikowski and Styczynsk [93], which
variations compared to other thinefilm modules, as their spectral reports any degradation after 6 years of operation connected to a
response has a much wider bandwidth. For example, in Ref. [84] solar inverter. However, there are survey reports that state much
there was almost no variation detected in the shortecircuit current higher averaged annual degradation rates for CIGS modules, such as
of these modules when plotted against the AM value. However, 1.17% or 1.28% [29,30].

864

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

Fig. 4. Evolution of main electrical parameters and parasitic resistances for module M3 (CdS/CdTe).

4. Conclusions other published studies. Having high annual degradation for ISC
(1.0%/year), the degradation rate in terms of VOC is significantly
A degradation analysis has been performed from outdoor lower (0.3%/year). However, RS has suffered the highest increment
measurements of four thinefilm photovoltaic modules that have of all the modules (23%/year), in addition to a significant but lower
been exposed for 20 months. Each specimen has been character- degradation of RSh (7.5%/year).
ized at regular intervals of time in such a way it has been possible to For the module based on CdTe/CdS technology, contrary to
perform a selection of samples within a very narrow range of modules based on amorphous silicon, the evolution of VOC and Vmax
irradiance and cell temperature to plot the evolution of the elec- is opposite to ISC and Imax; thus, the voltage values are the highest
trical parameters and of the parasitic resistances. The results show when APE is the lowest, but with a reduced amplitude of the sea-
a strong seasonal variation due to changes in the APE of the sun sonal trend in comparison to the other specimens. The obtained
spectral distribution and other phenomena as annealing. To annual degradation rate is also very significant (3.4%/year) for the
distinguish this annual variation from the longeterm degradation, CdS/CdTe module, due mainly to very high degradation of both
each filtered set of samples has been fitted to a novel degradation parasitic resistances, an increment of RS around 13%/year, and a
model, which allows to estimate the annual degradation rates of strong decrease of RSh reaching 26%/year.
the different indicators for each specimen. Finally, as for the Solyndra module (CIGS), despite of all the
Regarding the a-Si:H technology, both ISC and VOC are synchro- difficulties related to the very different features of this special de-
nized to the variation of the APE estimated from the spectrum. In vice, an estimation of its annual degradation rate in terms of power
this case, the highest annual degradation rate in terms of power has has been determined. Again, its VOC behavior is reversed with
been estimated (4.0%/year), among all the PV modules analyzed in respect the one observed for ISC. The most impacting fact about this
this paper, that could be due to a combination of the strongest specimen is the decrease of RS throughout time that, in addition to a
decrease of ISC (1.3%/year) and VOC (1.0%/year), in addition to a very low degradation of VOC and ISC, leads us to an insignificant
significant increase of RS (8.6%/year). Although there is a great annual degradation rate (0.2%/year); that is in agreement with the
dispersion of the degradation rates reported previously, the results results presented by some other authors.
presented in the paper are within the upper range stated in the Some final considerations should be made at the end. First, the
literature referred to a-Si:H. length of the exposure stabilization period is not a point clearly
With respect the aeSi/mceSi specimen, again the main electrical established in previous literature, and it is an open point of dis-
parameters follow the tendency of the APE evolution, but in the cussion. In any case, this issue could be an important source of
case of VOC and Pmax with more delay than in the case of a-Si:H. The discrepancy of the results obtained by different authors. Another
annual degradation rate in terms of power is also significantly high controversial element is the total length of the exposure period;
(3.4%/year), especially if it is compared to the one documented in thus, a longer one could lead to more representative longeterm

865

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

Fig. 5. Evolution of main electrical parameters and parasitic resistances for module M4 (CIGS).

Table 4 Table 6
Parameters Ki and degradation rates of M1. Parameters Ki and degradation rates of module M3.

K0 K1 K2 K3 DR K0 K1 K2 K3 DR

ISC 0.907 A
0.012 A/year
4.75e-2 þ0.190 years ¡1.3%/year ISC 0.962 A
0.004 A/year
2.26e-2 þ0.242 years ¡0.4%/year
VOC 82.31 V
0.81 V/year
5.51e-3 þ0.046 years ¡1.0%/year VOC 82.64 V
0.56 V/year þ5.62e-3 þ0.063 years ¡0.7%/year
Imax 0.723 A
0.016 A/year
5.45e-2 þ0.171 years ¡2.2%/year Imax 0.827 A
0.015 A/year
2.00e-2
0.218 years ¡1.8%/year
Vmax 61.19 V
1.15 V/year
1.27e-2 þ0.055 years ¡1.9%/year Vmax 61.17 V
1.07 V/year þ8.37e-3 þ0.084 years ¡1.7%/year
Pmax 44.22 W
1.77 W/year
6.42e-2 þ0.151 years ¡4.0%/year Pmax 50.60 W
1.75 W/year þ1.90e-2
0.149 years ¡3.5%/year
FF 59.23%
1.05%/year
1.67e-2 þ0.066 years ¡1.8%/year FF 63.68%
1.55%/year þ8.86e-3 þ0.096 years ¡2.4%/year
RS 3.177 U þ0.272 U/year
2.79e-2
0.013 years þ8.6%/year Rs 7.663 U 1.031 U/year þ2.32e-2
0.248 years þ13%/year
RSh 728.6 U
19.7 U/year
3.55e-2
0.181 years ¡2.7%/year RSh 1766.0 U
457.1 U/year þ1.31e-1 þ0.078 years ¡26%/year

Table 5 Table 7
Parameters Ki and degradation rates for module M2. Parameters Ki and degradation rates of module M4.

K0 K1 K2 K3 DR K0 K1 K2 K3 DR

ISC 2.601 A
0.025 A/year
2.53e-2 þ0.226 years ¡1.0%/year ISC 1.766 A
0.005 A/year þ6.21e-2
0.208 years ¡0.3%/years
VOC 54.57 V
0.18 V/year
1.40e-3
0.147 years ¡0.3%/year VOC 85.16 V
0.09 V/year þ6.63e-3 þ0.108 years ¡0.1%/years
Imax 2.100 A
0.045 A/year
2.99e-2 þ0.190 years ¡2.1%/year Imax 1.454 A
0.007 A/year þ6.22e-3
0.207 years ¡0.5%/years
Vmax 41.52 V
0.56 V/year
4.87e-3
0.032 years ¡1.3%/year Vmax 62.33 V þ0.20 V/year
7.65e-3
0.315 years þ0.3%/years
Pmax 87.20 W
3.00 W/year
3.10e-2 þ0.166 years ¡3.4%/year Pmax 90.60 W
0.150 W/year þ5.67e-2
0.193 years ¡0.2%/years
FF 61.43%
1.33%/year
1.14e-2 þ0.043 years ¡2.2%/year FF 60.23% þ0.16%/year
3.27e-3
0.178 years þ0.3%/years
RS 0.722 U 0.165 U/year
3.30e-2
0.166 years þ23%/year RS 3.401 U
0.270 U/year þ5.63e-2 þ0.1186 years ¡7.9%/years
RSh 163.4 U
12.2 U/year
3.54e-2 þ0.017 years ¡7.5%/year RSh 561.3 U
14.0 U/year
4.05e-2
0.252 years ¡2.5%/years

degradation rates. Finally, the posteprocessing of the measured The study presented in this article is focused on the degradation
data and the model used to remove the seasonal trend could have of thinefilm technologies at module level. It would be desirable to
an important impact on the final results; thus, it is necessary to extend the scope in order to analyze the power drop associated to
avoid data manipulation or correction procedures as far as possible, complete PV generators. Unfortunately, the scarcity of installations
as it has been done in this paper. based on these technologies and the difficulty to have the required

866

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

permissions and access to the necessary data remain challenges to conditions, Sol. Energy 157 (2017) 587e595, https://doi.org/10.1016/j.sol-
ener.2017.08.048. ISSN 0038-092X.
be overcome.
[12] S. Kichou, S. Silvestre, G. Nofuentes, M. Torres-Ramírez, A. Chouder, D. Guasch,
Characterization of degradation and evaluation of model parameters of
Funding amorphous silicon photovoltaic modules under outdoor long term exposure,
Energy 96 (2016) 231e241, https://doi.org/10.1016/j.energy.2015.12.054.
ISSN 0360-5442.
 e della Ricerca (Italy)
Ministero dell’Istruzione, dell’Universita [13] S. Kichou, E. Abaslioglu, S. Silvestre, G. Nofuentes, M. Torres-Ramírez,
[grant PRIN2020eHOTSPHOT 2020LB9TBC and grant A. Chouder, Study of degradation and evaluation of model parameters of
PRIN2017eHEROGRIDS 2017WA5ZT3_003]; Universita  degli Studi di micromorph silicon photovoltaic modules under outdoor long term exposure
in Jae
n, Spain, Energy Convers. Manag. 120 (2016) 109e119, https://doi.org/
Salerno [FARB funds]; Ministerio de Ciencia, Innovacio
n y Uni- 10.1016/j.enconman.2016.04.093. ISSN 0196-8904.
versidades (Spain) [grant RTI2018-095097eBeI0]. [14] S. Silvestre, S. Kichou, L. Guglielminotti, G. Nofuentes, M. Alonso-Abella,
Degradation analysis of thin film photovoltaic modules under outdoor long
term exposure in Spanish continental climate conditions, Sol. Energy 139
CRediT authorship contribution statement (2016) 599e607, https://doi.org/10.1016/j.solener.2016.10.030. ISSN 0038-
092X.
Michel Piliougine: Conceptualization, Methodology, Software, [15] P. Ferrada, F. Araya, A. Marzo, E. Fuentealba, Performance analysis of photo-
voltaic systems of two different technologies in a coastal desert climate zone
Validation, Data curation, Writing e original draft, Writing e re- of Chile, Sol. Energy 114 (2015) 356e363, https://doi.org/10.1016/j.sol-
view & editing, Visualization. Paula Sa
nchez-Friera: Methodology, ener.2015.02.009. ISSN 0038-092X.
Validation, Writing e original draft, Writing e review & editing. [16] A. Kyprianou, A. Phinikarides, G. Makrides, G.E. Georghiou, Definition and
computation of the degradation rates of photovoltaic systems of different
Giovanni Petrone: Methodology, Writing e original draft, Writing technologies with robust principal component analysis, IEEE J. Photovoltaics 5
e review & editing. Francisco Jose
Sa

nchez-Pacheco: Methodol- (6) (2015) 1698e1705, https://doi.org/10.1109/JPHOTOV.2015.2478065.
ogy, Writing e original draft, Writing e review & editing. Giovanni [17] V. Sharma, O. Sastry, A. Kumar, B. Bora, S. Chandel, Degradation analysis of
aeSi, (HIT) hetroejunction intrinsic thin layer silicon and meCeSi solar
Spagnuolo: Methodology, Writing e original draft, Writing e re- photovoltaic technologies under outdoor conditions, Energy 72 (2014)
view & editing. Mariano Sidrach-de-Cardona: Methodology, 536e546, https://doi.org/10.1016/j.energy.2014.05.078. ISSN 0360-5442.
Writing e original draft, Writing e review & editing. [18] J.Y. Ye, T. Reindl, A.G. Aberle, T.M. Walsh, Performance degradation of various
PV module technologies in tropical Singapore, IEEE J. Photovoltaics 4 (5)
(2014) 1288e1294, https://doi.org/10.1109/JPHOTOV.2014.2338051.
Declaration of competing interest [19] G. Makrides, B. Zinsser, M. Schubert, G.E. Georghiou, Performance loss rate of
twelve photovoltaic technologies under field conditions using statistical
techniques, Sol. Energy 103 (2014) 28e42, https://doi.org/10.1016/j.sol-
The authors declare that they have no known competing ener.2014.02.011. ISSN 0038-092X.
financial interests or personal relationships that could have [20] R. Gottschalg, T. Betts, A. Eeles, S. Williams, J. Zhu, Influences on the energy
appeared to influence the work reported in this paper. delivery of thin film photovoltaic modules, Sol. Energy Mater. Sol. Cell. 119
(2013) 169e180, https://doi.org/10.1016/j.solmat.2013.06.011. ISSN 0927-
0248.
References [21] N. Strevel, L. Trippel, M. Gloeckler, Performance characterization and superior
energy yield of First Solar PV power plants in high-temperature conditions,
[1] A. Jaeger-Waldau, in: PV Status Report 2010, Office for Official Publications of Photovoltaics International, 17, https://www.pv-tech.org/technical-papers/
the European Union, Luxembourg, 2010, https://doi.org/10.2788/87966. EUR performance-characterization-and-superior-energy-yield-of-first-solar-pv-
24344. power-plants-in-hightemperature-conditions/.
[2] L.V. Mercaldo, M.L. Addonizio, M.D. Noce, P.D. Veneri, A. Scognamiglio, [22] D.C. Jordan, S.R. Kurtz, Thin-film reliability trends toward improved stability,
C. Privato, Thin film silicon photovoltaics: architectural perspectives and in: 37th IEEE Photovoltaic Specialists Conference (PVSC), 2011, pp. 827e832,
technological issues, Appl. Energy 86 (10) (2009) 1836e1844, https://doi.org/ https://doi.org/10.1109/PVSC.2011.6186081. Seattle, WA, USA.
10.1016/j.apenergy.2008.11.034. ISSN 0306-2619,. [23] J. Del Cueto, S. Rummel, B. Kroposki, C. Osterwald, A. Anderberg, Stability of
[3] E.H. Rusnindyo, E.A. Karuniawan, A.A. Setiawan, F.D. Wijaya, Building inte- CIS/CIGS modules at the outdoor test facility over two decades, in: 33rd IEEE
grated thin film photovoltaic performance modelling on conventional build- Photovoltaic Specialists Conference (PVSC), 2008, pp. 1e6, https://doi.org/
ing, AIP Conf. Proc. 2255 (1) (2020), 070023, https://doi.org/10.1063/ 10.1109/PVSC.2008.4922772. San Diego, CA, USA.
5.0014597. [24] W. Durisch, K.-H. Lam, J. Close, Efficiency and degradation of a copper indium
[4] D.A. Cassini, S.C.S. Costa, A.S.A. Diniz, L.L. Kazmerski, Evaluation of failure diselenide photovoltaic module and yearly output at a sunny site in Jordan,
modes for photovoltaic modules in arid climatic zones in Brazil, in: 48th IEEE Appl. Energy 83 (12) (2006) 1339e1350, https://doi.org/10.1016/j.ape-
Photovoltaic Specialists Conference (PVSC), 2021, pp. 580e582, https:// nergy.2006.02.002. ISSN 0306-2619.
doi.org/10.1109/PVSC43889.2021.9518798. [25] J. Adelstein, W. Sekulic, Small PV systems PerformanceEvaluation at NREL's
[5] M. Kumar, A. Kumar, R. Gupta, Comparative degradation analysis of different OutdoorTest facility using the PVUSA power rating method, in: 2005 DOE
photovoltaic technologies on experimentally simulated water bodies and Solar Energy TechnologiesProgram Review Meeting, NREL/CP-520-39135,
estimation of evaporation loss reduction, Prog Photovoltaics 29 (3) (2021) 2005. Denver, CO, USA, https://www.osti.gov/biblio/882794-small-pv-sys-
357e378, https://doi.org/10.1002/pip.3370. tems-performance-evaluation-nrel-outdoor-test-facility-using-pvusa-power-
[6] R. Singh, M. Sharma, R. Rawat, C. Banerjee, Field Analysis of three different rating-method.
silicon-based Technologies in Composite Climate Condition e Part II e sea- [26] B. Marion, J. del Cueto, P. McNutt, D. Rose, Performance summary for the first
sonal assessment and performance degradation rates using statistical tools, solar CdTe 1ekW system, in: 2001 NCPV Program Review Meeting, NREL/CP-
Renew. Energy 147 (2020) 2102e2117, https://doi.org/10.1016/ 520-30942, National Renewable Energy Laboratory NREL, Golden, CO, USA,
j.renene.2019.10.015. ISSN 0960-1481. 2001. https://www.osti.gov/biblio/15000007.
[7] S. Kichou, P. Wolf, S. Silvestre, A. Chouder, Analysis of the behaviour of cad- [27] D.C. Jordan, T.J. Silverman, J.H. Wohlgemuth, S.R. Kurtz, K.T. VanSant, Photo-
mium telluride and crystalline silicon photovoltaic modules deployed outdoor voltaic failure and degradation modes, Prog Photovoltaics 25 (4) (2017)
under humid continental climate conditions, Sol. Energy 171 (2018) 681e691, 318e326, https://doi.org/10.1002/pip.2866.
https://doi.org/10.1016/j.solener.2018.07.028. ISSN 0038-092X. [28] D.C. Jordan, S.R. Kurtz, K. VanSant, J. Newmiller, Compendium of photovoltaic
[8] N. Aarich, M. Raoufi, A. Bennouna, N. Erraissi, Outdoor comparison of rooftop degradation rates, Prog Photovoltaics 24 (7) (2016) 978e989, https://doi.org/
grideconnected photovoltaic technologies in Marrakech (Morocco), Energy 10.1002/pip.2744.
Build. 173 (2018) 138e149, https://doi.org/10.1016/j.enbuild.2018.05.030. [29] S. Chattopadhyay, R. Dubey, V. Kuthanazhi, S. Zachariah, S. Bhaduri,
ISSN 0378-7788. C. Mahapatra, et al., All-India Survey of Module Reliability: 2016, National
[9] A. Limmanee, S. Songtrai, N. Udomdachanut, S. Kaewniyompanit, Y. Sato, Institute of Solar Energy (India) (2017). https://nise.res.in/wp-content/
M. Nakaishi, S. Kittisontirak, K. Sriprapha, Y. Sakamoto, Degradation analysis uploads/2018/01/All-India-Survey-of-Photovoltaic-Module-Reliability-
of photovoltaic modules under tropical climatic conditions and its impacts on 2016-_-Rev2_25012018_online_lowres.pdf.
LCOE, Renew. Energy 102 (2017) 199e204, https://doi.org/10.1016/ [30] Y. Golive, S. Zachariah, S. Bhaduri, R. Dubey, S. Chattopadhyay, P. Joshi, et al.,
j.renene.2016.10.052. ISSN 0960-1481. in: All-India Survey of Module Reliability: 2018, National Institute of Solar
[10] T. Ozden, B.G. Akinoglu, R. Turan, Long term outdoor performances of three Energy, 2019 (India), https://nise.res.in/wp-content/uploads/2020/01/All-
different onegrid PV arrays in central Anatolia e an extended analysis, India-Survey-of-Photovoltaic-Module-Reliability_2018_Report.pdf.
Renew. Energy 101 (2017) 182e195, https://doi.org/10.1016/ [31] J.A. Del Cueto, S. Rummel, B. Kroposki, A. Anderberg, Long-term performance
j.renene.2016.08.045. ISSN 0960-1481. data and analysis of CIS/CIGS modules deployed outdoors, in: B. von Roedern,
[11] A. Tahri, S. Silvestre, F. Tahri, S. Benlebna, A. Chouder, Analysis of thin film A.E. Delahoy (Eds.), Photovoltaic Cell and Module Technologies II, vol. 7045,
photovoltaic modules under outdoor long term exposure in semi-arid climate International Society for Optics and Photonics, SPIE, San Diego, CA, USA, 2008,

867

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

pp. 7e17, https://doi.org/10.1117/12.796505. [53] Kaneka GEA060 specification sheet, KANEKA. https://www.dropbox.com/s/
[32] T. Silverman, U. Jahn, G. Friesen, Characterisation of performance of thinefilm mrr7xm5qeyi1iaf/Kaneka_GEA060.pdf. November 15th, 2021, Corporation.
photovoltaic technologies: IEA PVPS task 13, subtask 3.1 : final report [54] Solarmodule Phoenix solar e PHX e 120 e LV, Phoenix solar. https://www.
IEAePVPS T13-02:2014, international energy agency, photovoltaic power dropbox.com/s/4ohvib18ink1hqa/PhoenixSolar_PHX120LV.pdf. November
systems programme, I. E. A. P. P. S. Programme, https://repository.supsi.ch/ 15th, 2021, AG.
5776/1/IEA-PVPS_T13-02_2014_Characterization_ThinFilm_Modules.pdf, [55] First Solar FS Series 2 PV Module, PD-5-401-02, First Solar Inc, 2011. https://
2014, 978-3-906042-17-6. www.dropbox.com/s/xmmny6vjzifqo60/FirstSolar.pdf. (Accessed 15
[33] D. Dirnberger, Uncertainty in PV module measurement e Part II: verification November 2021).
of rated power and stability problems, IEEE J. Photovoltaics 4 (3) (2014) [56] Solar modules SOLYNDRA-SL-001-150C/157C/165C/173C/182C, Specification
991e1007, https://doi.org/10.1109/JPHOTOV.2014.2307158. sheet, sly182-eng-01109, Phoenix solar. https://www.dropbox.com/s/
[34] D.L. Staebler, C.R. Wronski, Reversible conductivity changes in d3c7n22y90g4red/Solyndra.pdf. November 15th, 2021, Inc.
dischargeeproduced amorphous Si, Appl. Phys. Lett. 31 (4) (1977) 292e294, [57] Solyndra product training, Solyndra, URL, https://www.dropbox.com/s/
https://doi.org/10.1063/1.89674. si3ozo9rxk5o1wg/SolyndraManual.pdf. November 15th, 2021, Inc.
[35] T. Yanagisawa, Longeterm degradation tests of amorphous silicon solar cells: [58] Precision Pyranometer CM21 e Instruction Manual, Kipp & Zonen, Delft,
correlation between lighte and currenteinduced degradation characteristics, Holland, 0304 edn, https://www.dropbox.com/s/hln7wqhoakb2ax6/
Microelectron. Reliab. 35 (2) (1995) 183e187, https://doi.org/10.1016/0026- KIPPZONEN_cm21.pdf. (Accessed 15 November 2021) (B.V).
2714(95)90084-4. ISSN 0026-2714. [59] Compensated calibrated cell e radiation sensors, Atersa, Grupo elecnor,
[36] Kaneka Thin Film PV Installation Manual, in: Module Type: G-Ea060, KANEKA Madrid, espan ~ a. https://www.dropbox.com/s/khmn20h9em450d9/atersa.pdf,
Corporation, 2009. https://www.dropbox.com/s/8n9w6lc077nhcso/KANEKA_ 2004. (Accessed 15 November 2021).
MANUAL.pdf. (Accessed 15 November 2021). [60] Adhesive silicone patch Pt100 temperature sensors, Industrial Temperature
[37] M. Mun ~ ozeGarcía, O. Marin, M. AlonsoeGarcía, F. Chenlo, Characterization of Sensors, Last accessed: November 15th, 2021, ITS, https://www.dropbox.com/
thin film PV modules under standard test conditions: results of indoor and s/ae41d240hkw72r3/pt100.pdf.
outdoor measurements and the effects of sunlight exposure, Sol. Energy 86 [61] L. Micheli, J.A. Caballero, E.F. Fern
andez, G.P. Smestad, G. Nofuentes,
(10) (2012) 3049e3056, https://doi.org/10.1016/j.solener.2012.07.015. ISSN T.K. Mallick, F. Almonacid, Correlating photovoltaic soiling losses to waveband
0038-092X. and single-value transmittance measurements, Energy 180 (2019) 376e386,
[38] C. Radue, E. Van Dyk, A comparison of degradation in three amorphous silicon https://doi.org/10.1016/j.energy.2019.05.097. ISSN 0360-5442.
PV module technologies, Sol. Energy Mater. Sol. Cell. 94 (3) (2010) 617e622, [62] IEC 60904e3, in: Photovoltaic Devices e Part 3: Measurement Principles for
https://doi.org/10.1016/j.solmat.2009.12.009. ISSN 0927-0248. Terrestrial Photovoltaic (PV) Solar Devices with Reference Spectral Irradiance
[39] J.A. Del Cueto, C.A. Deline, S.R. Rummel, A. Anderberg, Progress toward a Data, 4 edn., International Electrotechnical Commission IEC, Geneva,
stabilization and preconditioning protocol for polycrystalline thinefilm Switzerland, 2019, ISBN 978-2-8322-6268-9. https://webstore.iec.ch/
photovoltaic modules, in: 35th IEEE Photovoltaic Specialists Conference publication/61084.
(PVSC), 2010, pp. 2423e2428, https://doi.org/10.1109/PVSC.2010.5614223. [63] Compact FieldPoint cFP-21xx and cFP-BP-X e User Manual, National In-
Honolulu, HI, USA. struments, Austin (TX, USA), 2005. https://www.dropbox.com/s/
[40] K. Astawa, T. Betts, R. Gottschalg, Effect of loading on long term performance 7gxfnw8zll1ksun/cFP_2120.pdf. (Accessed 15 November 2021).
of single junction amorphous silicon modules, Sol. Energy Mater. Sol. Cell. 95 [64] M. Piliougine, J. Carretero, L. Mora-Lo
pez, M. Sidrach-de-Cardona, Experi-
(1) (2011) 119e122, https://doi.org/10.1016/j.solmat.2010.04.071. ISSN 0927- mental system for currentevoltage curve measurement of photovoltaic
0248. modules under outdoor conditions, Prog Photovoltaics 19 (5) (2011)
[41] D. King, J. Kratochvil, W. Boyson, Stabilization and performance characteristics 591e602, https://doi.org/10.1002/pip.1073.
of commercial amorphousesilicon PV modules, in: 28th IEEEE Photovoltaic [65] M. Piliougine, J. Carretero, L. Mora-Lo
pez, M. Sidrach-de-Cardona, New soft-
Specialists Conference (PVSC), 2000, pp. 1446e1449, https://doi.org/10.1109/ ware tool to characterize photovoltaic modules from commercial equipment,
PVSC.2000.916165. Anchorage, AK, USA. WEENTECH Proceedings in Energy (2018) 211e220, https://doi.org/10.32438/
[42] J.A. Del Cueto, J. Pruett, D. Cunningham, Stabilization of high efficiency CdTe WPE.6218.
photovoltaic modules in controlled indoor light soaking, in: National Center [66] M. Piliougine, A. Oukaja, P. S

nchez-Pacheco,
anchez-Friera, G. Petrone, F.J. Sa
for Photovoltaics and Solar Program Review Meeting, 2003. NREL/CP-520- G. Spagnuolo, M. Sidrach-de-Cardona, Analysis of the degradation of single-
33543, Denver, CO, USA, https://www.osti.gov/biblio/15004239-stabilization- crystalline silicon modules after 21 years of operation, Prog Photovoltaics
high-efficiency-cdte-photovoltaic-modules-controlled-indoor-light-soaking. 29 (8) (2021) 907e919, https://doi.org/10.1002/pip.3409.
[43] C. Deline, J.A. del Cueto, D.S. Albin, S. Rummel, Metastable electrical charac- [67] ISO/IEC Guide 98-3, in: Uncertainty of Measurement e Part 3: Guide to the
teristics of polycrystalline thin-film photovoltaic modules upon exposure and Expression of Uncertainty in Measurement (GUM:1995), 1 edn., International
stabilization, J. Photon. Energy 2 (1) (2012), 022001, https://doi.org/10.1117/ Organization for Standardization ISO, Geneva (Switzerland), 2008. https://
1.JPE.2.022001. www.iso.org/standard/50461.html.
[44] Resumen Mensual de la radiacio
n solar. Diciembre 2011, Agencia Estatal de [68] K. Emery, in: Photovoltaic Calibrations at the National Renewable Energy
Meteorología (AEMET). Ministerio de Medio Ambiente y Medio Rural y Mar- Laboratory and Uncertainty Analysis Following the ISO 17025 Guidelines,
ino. Gobierno de Espan ~ a. https://www.aemet.es/documentos/es/ National Renewable Energy Laboratory NREL, Golden, CO, USA, 2016. NREL/
serviciosclimaticos/vigilancia_clima/radiacion_ozono/radiacion_solar/2011/ TP-5J00-66873, https://www.nrel.gov/docs/fy17osti/66873.pdf.
InformeRad_Solar_2011_12.pdf, 2011. (Accessed 15 November 2021). [69] F. Toledo, J.M. Blanes, Analytical and quasi-explicit four arbitrary point
[45] B.H. King, J.S. Stein, D. Riley, C.B. Jones, C.D. Robinson, Degradation assessment method for extraction of solar cell single-diode model parameters, Renew.
of fielded CIGS photovoltaic arrays, in: 44th IEEE Photovoltaic Specialist Energy 92 (2016) 346e356, https://doi.org/10.1016/j.renene.2016.02.012.
Conference (PVSC), 2017, https://doi.org/10.1109/PVSC.2017.8366332. ISSN 0960-1481.
Washington, DC, USA, 3155e3160. [70] M. Piliougine, D. Elizondo, L. Mora-Lo
pez, M.S. de Cardona, Photovoltaic
[46] A. Kaul, S. Pethe, N.G. Dhere, Long-term performance analysis of copper in- module simulation by neural networks using solar spectral distribution, Prog
dium gallium selenide thin-film photovoltaic modules, J. Photon. Energy 2 (1) Photovoltaics 21 (5) (2013) 1222e1235, https://doi.org/10.1002/pip.2209.
(2012), 022005, https://doi.org/10.1117/1.JPE.2.022005. [71] R. MorenoeSa
ez, M. SidrachedeeCardona, L. MoraeLo
pez, Analysis and
€
[47] M. Demirtaş, B. Tamyürek, E. Kurt,
Ipek Çetinbaş, M.K. Oztürk, Effects of aging characterization of photovoltaic modules of three different thin-film tech-
and environmental factors on performance of CdTe and CIS thinefilm nologies in outdoor conditions, Appl. Energy 162 (2016) 827e838, https://
photovoltaic modules, J. Electron. Mater. 48 (11) (2019) 6890e6900, https:// doi.org/10.1016/j.apenergy.2015.10.156. ISSN 0306-2619,.
doi.org/10.1007/s11664-019-07172-z. ISSN 1543-186X. [72] G. Nofuentes, C. Gueymard, J. Aguilera, M. Pe
rez-Godoy, F. Charte, Is the
[48] M. Kumar, A. Kumar, Experimental validation of performance and degradation average photon energy a unique characteristic of the spectral distribution of
study of canal-top photovoltaic system, Appl. Energy 243 (2019) 102e118, global irradiance? Sol. Energy 149 (2017) 32e43, https://doi.org/10.1016/
https://doi.org/10.1016/j.apenergy.2019.03.168. ISSN 0306-2619. j.solener.2017.03.086. ISSN 0038-092X.
[49] A. Virtuani, H. M ullejans, E.D. Dunlop, Comparison of indoor and outdoor [73] S. Lindig, I. Kaaya, K.-A. Weiss, D. Moser, M. Topic, Review of statistical and
performance measurements of recent commercially available solar modules, analytical degradation models for photovoltaic modules and systems as well
Prog Photovoltaics 19 (1) (2011) 11e20, https://doi.org/10.1002/pip.977. as related improvements, IEEE J. Photovoltaics 8 (6) (2018) 1773e1786,
[50] IEC 60891, in: Photovoltaic Devices e Procedures for Temperature and Irra- https://doi.org/10.1109/JPHOTOV.2018.2870532.
diance Corrections to Measured IeV Characteristics, 3 edn., International [74] R. Rüther, J. Livingstone, Seasonal variations in amorphous silicon solar
Electrotechnical Commission IEC, Geneva (Switzerland), 2021, 978-2-8322- module outputs and thin film characteristics, Sol. Energy Mater. Sol. Cell. 36
1036-0, 2021, https://webstore.iec.ch/publication/61766. (1) (1995) 29e43, https://doi.org/10.1016/0927-0248(94)00165-O. ISSN
[51] M. Piliougine, R. Guejia-Burbano, G. Petrone, F. S
anchez-Pacheco, L. Mora- 0927-0248.
Lo
pez, M. Sidrach-de-Cardona, Parameters extraction of single diode model [75] P. Ragot, H. Costa, D. Desmettre, Higher efficiency new generation amorphous
for degraded photovoltaic modules, Renew. Energy 164 (2021) 674e686, silicon PV modules, in: 11th European Photovoltaic Solar Energy Conference,
https://doi.org/10.1016/j.renene.2020.09.035. ISSN 0960-1481. Montreux, Switzerland, 1992, pp. 553e555.
[52] E.A. Schiff, S. Hegedus, X. Deng, Amorphous siliconebased solar cells, chap. 12, [76] G. Kleiss, K. Bücher, A. Raicu, K. Heidler, Performance of thin film solar cells
in: A. Luque, S. Hegedus (Eds.), Handbook of Photovoltaic Science and Engi- under realistic reporting conditions including degradation effects, in: 11th
neering, 2 edn, John Wiley and Sons, Chichester, West Sussex, UK, 2011, European Photovoltaic Solar Energy Conference, Montreux, Switzerland,
https://doi.org/10.1002/9780470974704.ch12, 978-0-470-72169-8, 487e545, 1992, pp. 578e581.
doi:10.1002/9780470974704.ch12, 978-0-470-72169-8. [77] P. Ragot, J.L. Martin, Performance of thin film solar cells under realistic

868

nchez-Friera, G. Petrone et al.
M. Piliougine, P. Sa Renewable Energy 193 (2022) 857e869

reporting conditions including degradation effects, in: 13th European module fill factors, Sol. Energy Mater. Sol. Cells 71 (3) (2002) 375e385,
Photovoltaic Solar Energy Conference, 1995. Nice, France, 2315e2318. https://doi.org/10.1016/S0927-0248(01)00095-2.
[78] A. Carr, T. Pryor, A comparison of the performance of different PV module [87] L. Fanni, I. Pola, E. Bur
a, T. Friesen, D. Chianese, Investigation of annealing and
types in temperate climates, Sol. Energy 76 (1) (2004) 285e294, https:// degradation effects on a-Si PV modules in real operating conditions, in: 24th
doi.org/10.1016/j.solener.2003.07.026. ISSN 0038-092X. European Photovoltaic Solar Energy Conference, EU PVSEC, Hamburg, Ger-
[79] M. Pierro, F. Bucci, C. Cornaro, Impact of light soaking and thermal annealing many, 2009, https://doi.org/10.4229/24thEUPVSEC2009-4AV.3.92,
on amorphous silicon thin film performance, Prog Photovoltaics 23 (11) 3596e3599.
(2015) 1581e1596, https://doi.org/10.1002/pip.2595. [88] D. Jordan, S.R. Kurtz, Photovoltaic degradation rates e an analytical review,
[80] A. Phinikarides, G. Makrides, N. Kindyni, A. Kyprianou, G.E. Georghiou, ARIMA Prog Photovoltaics 21 (1) (2013) 12e29, https://doi.org/10.1002/pip.1182.
modeling of the performance of different photovoltaic technologies, in: 39th [89] J.A. Del Cueto, Review of the field performance of one cadmium telluride
IEEE Photovoltaic Specialists Conference (PVSC), 2013, https://doi.org/ module, Prog Photovoltaics 6 (6) (1998) 433e446, https://doi.org/10.1002/
10.1109/PVSC.2013.6744268. Tampa, FL, USA, 0797e0801. (SICI)1099-159X(199811/12)6:6<433::AID-PIP236>3.0.CO;2-L.
[81] S.A. Tabatabaei, D. Formolo, J. Treur, Analysis of performance degradation of [90] L. Ngan, N. Strevel, K. Passow, A.F. Panchula, D. Jordan, Performance charac-
domestic monocrystalline photovoltaic systems for a realeworld case, Energy terization of cadmium telluride modules validated by utility-scale and test
Proc. 128 (2017) 121e129, https://doi.org/10.1016/j.egypro.2017.09.025. systems, in: 40th IEEE Photovoltaic Specialist Conference (PVSC), 2014,
[82] I. Muirhead, B. Hawkins, Research into new technology photovoltaic modules https://doi.org/10.1109/PVSC.2014.6925309. Denver, CO, USA, 1957e1962.
at Telstra Research Laboratories e what we have learnt, in: Annual Confer- [91] H. Hiraki, A. Hiraki, M. Maeda, Y. Takahashi, Unique features of cylindrical
ence of the Australian and New Zealand Solar Energy Society - SOLAR ’96, type solaremodule contrasted with plane or conventional type ones, J Phys
1996. Darwin, Northern Territory, Australia, https://www.dropbox.com/s/ Conf Ser 371 (1) (2012), 012002, https://doi.org/10.1088/1742-6596/379/1/
874wolo29tvfgov/Solar96.PDF. 012002.
[83] Matlab Optimization Toolbox. User's Guide. R2021b, MathWorks, Natick (MA, [92] C.M. Gronet, Solyndra, inc. Amendment No. 1 to Form S-1 registration state-
USA), 2021. https://www.dropbox.com/s/y9c49mj118hkd8l/optim.pdf. ment, reg. Number 333-163845. https://www.sec.gov/Archives/edgar/data/
[84] E. Meyer, E. van Dyk, Characterization of degradation in thin-film photovoltaic 0001443115/000119312510058567/ds1a.htm, 2010. (Accessed 15 November
module performance parameters, Renew. Energy 28 (9) (2003) 1455e1469, 2021).
https://doi.org/10.1016/S0960-1481(02)00062-9. [93] H.D. Musikowski, Z.A. Styczynsk, Analysis of the operational behavior and
[85] J. Merten, J. Andreu, Clear separation of seasonal effects on the performance of longeterm performance of a CIS PV system, in: 25th European Photovoltaic
amorphous silicon solar modules by outdoor IV measurements, Sol. Energy Solar Energy Conference and Exhibition (PVSEC)/5th World Conference on
Mater. Sol. Cell. 52 (1) (1998) 11e25, https://doi.org/10.1016/S0927-0248(97) Photovoltaic Energy Conversion, WCPEC), Valencia, Spain, 2010,
00263-8. ISSN 0927-0248. pp. 3942e3946, https://doi.org/10.4229/25thEUPVSEC2010-4DO.11.4.
[86] R. Rüther, G. Kleiss, K. Reiche, Spectral effects on amorphous silicon solar

869