Design and Build Technical Briefing

Understanding the energy

yield of PV modules
Module yield | Varying climatic conditions across markets and the individual characteristics of PV
technologies undermine accurate predictions of module energy yield using conventional methods.
Markus Schweiger, Werner Herrmann, Christos Monokroussos and Uwe Rau describe how a
calculation of module performance ratio can be used to accurately assess the efficiency of different
PV module technologies in different climates and thus the likely return on investment from a project

B
etween 2004 and 2016 a sum of yet [3,4]. The energy yield estimation for with a moderate specific energy yield of
US$1,161 billion was invested in PV various PV module technologies, using 1,500kWh/kWp and a levelised cost of
systems [1], and there is currently simulation tools, exhibits high uncertain- electricity (LCOE) of US$100/MWh; this
approximately 200GW of PV capacity ties as a result of the limited availability of means US$150,000 extra revenue for
installed worldwide. By 2050 a globally sufficient PV module performance data. each per cent of additional energy yield
installed PV capacity of around 4.6TWp It is therefore essential to have a and year of operation (if emerging inter-
is expected; this in turn implies a global detailed understanding of all the factors est earnings are neglected). This would
investment market of some US$225 billion that impact on the energy yield perfor- essentially mean US$3.75 million more
per year on average through 2050 [2]. mance of PV modules. Such knowledge revenue per 1% increase in energy yield
A major part of this investment is will provide a scientific basis for making after 25 years of operation. Furthermore,
represented by the price of PV modules, accurate yield estimates for different assuming a new market of around 4.4TWp
which is determined by their output technologies and for optimising energy as mentioned earlier, and while keeping
power rated at standard test conditions yield performance for different climates. the specific energy yield, lifetime and
(STC), specifically an irradiance of 1,000W/ For the upcoming multi-GW installations LCOE constant, the result is an astonishing
m2, a module temperature of 25°C and of 125GW/year on average, each percent- US$165 billion surplus in revenue per 1%
a spectral irradiance according to IEC age of uncertainty results in significant of energy yield, which could be achieved
60904-3. Real outdoor operating condi- investment uncertainty with regard to by choosing capable PV modules. Besides
tions, however, are in general consider- capital expenditures. the chance for investors to maximise their
ably different from STC conditions, as net profit by considering the energy yield
demonstrated in Figures 1 and 2 for Energy yield performance as a key performance, this relation also bears a
optimal mounting conditions. The relevant factor for the return on a PV invest- certain investment risk for the PV industry
standards for specifying the energy rating ment if the long-term performance is lower than
of PV modules are IEC 61853 parts 1 to Consider a PV power plant with 100MWp expected, and if investors are not able to
4, but not all parts have been published nominal power (for STC) at a location accurately calculate the expected income.

Figure 1. The test sites operated by TÜV Rheinland for PV module characterisation and energy yield measurements (clockwise from left): Cologne
(Germany, moderate climate), Tempe (Arizona, dry continental climate), Chennai (India, tropic climate), Thuwal (Saudi Arabia, dry desert climate with
sand deposition) and Ancona (Italy, Mediterranean climate).

56 | May 2017 | www.pv-tech.org

 Design and Build Technical Briefing

Figure 2. Generated electrical energy

of a crystalline PV module in five
different climates as a function of
module temperature and irradiance
on an annual basis, compared with the
measuring conditions of IEC 61853-1 The MPR is suitable for investigating
energy rating matrix (red dots). Colour the efficiency of PV modules in different
range: 0.1–2.6%; colour increment: 0.1%
climates compared with STC efficiency, as
well as for comparing different technolo-
gies and climates. As the local weather
From absolute yield to specific conditions cannot be changed (unlike the
yield to module performance ratio global climate), differences with respect
The energy yield of PV modules deployed to technological origin are of special
in different climates is a complex topic interest for optimising PV module perfor-
involving interdisciplinary knowledge mance and for selecting suitable products
of cell physics, module properties and for a certain climate. The amount by
meteorological aspects. To find a pathway which the value of MPR differs from unity
to the underlying correlations, some represents the losses in real outdoor
general definitions therefore need to be operating conditions compared with STC
discussed first. efficiency. The MPR facilitates a relative
The absolute energy yield (EY) of PV comparison in percentage terms between
modules is defined in watt hours (Wh). different technologies and climates; it
Because of the different efficiencies and includes all the offset relevant influences
designs of PV modules, it makes sense on energy yield performance due to
to calculate the specific energy yield in inaccurate nominal power, temperature
watt hours per watt peak (kWh/kWp), by losses, non-linear module performance
dividing EY by the nominal power PSTC; depending on irradiance G (low-irradiance
this allows a comparison of the energy behaviour) and spectral effects, as well
yield performances of different types as the losses due to soiling and angular
of PV module. Besides PSTC the second behaviour (as illustrated in Figure 3). The
factor dominating energy yield is solar MPR is identical to the performance ratio
irradiation (H); this strongly depends (PR), commonly used for PV systems,
on geographic location, local mounting when system losses, such as wiring,
conditions of the PV power plant and module mismatch or invertor losses, are
annual fluctuations. When choosing a not considered. Uncertainties of less than
pyranometer as a reference irradiance ±1% can be achieved when choosing
sensor, H is almost independent of PSTC as stated by the manufacturers as a
environment-related impact factors, such constant basis for MPR calculations.
as angle of incidence, spectral shifts or
temperature. Thus, to compare and elabo- Underlying database and investi-
rate only technology-driven performance gations performed
factors, the module performance ratio Since 2013 the performance of 15 differ-
(MPR) is the best-practice method and ent PV module types within the nation-
can be calculated as: ally founded ‘PVKlima’ R&D project has

Figure 3. Factors influencing the energy yield of PV modules

58 | May 2017 | www.pv-tech.org

 Design and Build Technical Briefing

The five different c-Si module types compared. The method allows the influ-
comprise three polycrystalline and one ence of temperature and spectral irradi-
monocrystalline PV modules with hetero- ance on fill factor FF, short-circuit current
junction cells, and one monocrystalline Isc, open-circuit voltage Voc and PSTC to be
module with back-contacted n-type cells. analysed independently of each other.
The polycrystalline samples are equipped Starting with c-Si, mostly stable PSTC
with different front glasses: one sample power values were found within more
with standard float glass, one with an anti- than three years of outdoor exposure for
reflection coating and one with deeply all climates. Typical long-term average
structured glass. degradation rates of less than –0.5% per
Comprehensive tests with regard to year can be confirmed. For heterojunction
energy rating and energy yield were PV modules, higher rates of about –1.0%
performed in the laboratory and outdoors; per year were observed, mainly related to
five test sites, each in a different climate
zone, were therefore constructed (see Fig.
1). The annual in-plane global solar irradia- “The most important
tion was 2,386kWh/m2 in Saudi Arabia,
2,360kWh/m2 in Arizona, 1,860kWh/m2 in
pieces of information
India, 1,556kWh/m2 in Italy and 1,195kWh/ for investors are the
m2 in Germany. These test sites allow the
generation of the PV module and environ- results based on the
mental data sets needed to understand
the real-world performance and long-term
pure STC power as
reliability of PV modules. Thus it was possi- stated and sold by the
ble to generate an understanding (that so
far is unique) of PV module performance manufacturers”
under real operating conditions in differ-
ent climates. a decrease in Voc. The c-Si 1 sample shown
in Figure 4 exhibits an approximately 4%
Nominal power at STC and lower value than the PSTC stated on the
monitoring of electrical stability label. It is noted that the stated results are
To understand the energy yield of PV subject to a measurement uncertainty of
modules, it is necessary to first begin with ±2.5%, which should be borne in mind
the most challenging aspect from the when interpreting the results.
metrology point of view: the determina- The nominal power of CIGS PV modules
tion of STC power and the monitoring of revealed significant performance changes
its stability during outdoor operation. due to metastable cell processes; the
To get a deeper insight into the various consolidation phase of these processes
seasonal effects on module performance, can take longer than a year. Changes in
an elaborate current–voltage (I–V) curve power are related in equal proportions to
analysis was employed. After the I–V FF and Voc. The depicted CIGS 2 sample
curves of all samples were measured using in Figure 4 exhibited an increasing PSTC
a sampling rate of 10 minutes, corrections of around +4% compared with the label
of temperature and irradiance according specification. Some of the other PV
to IEC 60891 [5] were applied, in combina- module types resulted in more than a
Figure 4. Monthly averages of STC-corrected nominal power tion with a spectral mismatch correction –10% deviation from the label value after
for four PV module types, normalised to stated nominal power, obtained from measured spectral irradi- three years of operation; this depends
for the Ancona test site in Italy (uncertainty: ±2.5%)
ance data according to IEC 60904-7 [6]. on the manufacturer and not just on the
These corrections are necessary in order technology.
been undergoing systematic analysis. The to create constant operating conditions The tested CdTe PV modules also
tested modules were: for time series analysis which would not revealed metastable processes that
• Five different crystalline silicon (c-Si) otherwise be achieved outdoors. significantly affected PSTC. After an initial
module types from three different Figure 4 shows the monthly average performance increase of up to 8%, which
manufacturers. STC power for four samples representing takes several months (depending on local
• Four Cu(In,Ga)Se2 (CIGS) modules from four technologies. The test site in Italy is temperature conditions), the nominal
four different manufacturers. used as a model case for the discussion of power exhibits annealing processes
• Three cadmium telluride (CdTe) module some fundamental PV module perfor- between summer and winter, leading
types from two different manufacturers. mance characteristics. to a PSTC oscillation with an amplitude
• Three amorphous silicon (a Si tandem) With the application of this correction of approximately ±2%; this oscillation
module variants from three different method, all environmental influences disappears in hot climates, such as those
manufacturers. are accounted for and can be directly found in Tempe or Chennai. The annealing

60 | May 2017 | www.pv-tech.org

 Design and Build Technical Briefing

just a few months of operation. Seasonal performance behaviour

It remains unanswered here whether or under investigation
not the technology-specific stabilisation To investigate the origin of the above-
procedures stated in the new IEC 61215 mentioned significant differences in
[7] series of standards are suitable in order annual yield results, an evaluation of
to achieve reliable, stabilised PSTC values. short-term MPR values provides a first
All the results on stability can be reviewed impression of the physical background.
in Schweiger et al. [8]. Now that the It is a fast and easy way to obtain insights
PSTC values of all PV modules have been into module performance, which is also
verified, the discussion about climate- the reason why it is used most frequently
related influences can continue. as a monitoring solution for PV systems,
needing just one reference irradiance
Origin of climate-related perfor- sensor. The potential, however, is limited,
mance differences for PV module since all influencing factors are superim-
technologies and major findings posed onto just a single value.
As mentioned above, PV modules have For the MPR calculation, the maximum
different low-irradiance behaviours, differ- power point was tracked with a sampling
ent temperature coefficients, different frequency of 30s, and a ventilated
operating temperatures, different spectral pyranometer served as a reference irradi-
and angular behaviours and also different ance sensor. Figure 5 shows the monthly
soiling behaviours when different front average MPR values of representative
glasses are used. These factors, combined samples in Italy based on stated PSTC,
with site-specific climate conditions, result together with the compensated MPR
in significant performance differences on based on measured PSTC, as well as
the basis of the nominal power measured temperature losses and spectral irradiance
at STC. As pointed out, the nominal power influences. This plot is used again as the
can deviate significantly up or down from model case for the discussion of some
Figure 5. Monthly module performance ratio based on stated
PSTC for four PV module types, compared with the MPR based the stated values as a result of binning fundamental performance characteristics
on monthly measured PSTC, for the Ancona test site in Italy. A policies, measuring inaccuracies (±2% in of different PV module technologies.
deviation from 100% means yield losses or gains; the tempera- the laboratory) or stability issues, such as As discussed earlier, the performance
ture (orange) and spectral effect (purple) contributions are
light-induced degradation (LID), potential- of c-Si PV modules (black dots, Figure 5)
indicated
induced degradation (PID), or metastabili- is mostly stable. Nevertheless, the plot of
ties for thin film. monthly MPR values for c-Si shows the
process is assumed to achieve a constant Given the impact on investment of just strongest oscillations by season, with
state in these hot locations for the whole one percentage point difference in energy maximal MPR values in winter; the reason
year. Changes in power are related mainly yield performance, the most important for this is the high relative tempera-
to FF. The average stabilised PSTC of the pieces of information for investors are ture coefficient γ, with typical values of
CdTe 1 sample shown in Figure 4 fits quite the results based on the pure STC power –0.35%/K for high-efficiency modules and
well with the stated PSTC values after three as stated and sold by the manufacturers. –0.42%/K for standard cells. The maximum
years of operation; however, PV module Within this project, a significant differ- in winter can be reduced for modules of
types with more than a –10% deviation ence in the energy yield performance was each technology with poor low-irradiance
from the label value after three years of observed between the best- and worst- behaviour due to the lower average irradi-
operation were also found. performing PV module types: up to 23% ances on winter days. The influences of
The performance of a-Si PV modules in India, 21% in Arizona, 14% in Germany spectral effects on c-Si are low. An offset of
revealed the well-known (but not fully and 12% in Italy. After compensating the MPR curves can occur in the case of PV
understood) Staebler–Wronski effect, the effects related to nominal power modules with inaccurately stated nominal
with initial stabilisation of around –10% to mismatch discussed earlier, an annual power on the label or datasheet.
–15%, depending on module type. As in difference in yield of 16% in India, 19% in Almost the same performance behav-
the case of CdTe, the performance reveals Arizona, 8% in Germany and 9% in Italy iour can be observed for CIGS samples
a summer and winter oscillation of about remained; the results for Saudi Arabia are (blue dots, Figure 5); the spectral response
±3%, which could also be observed for still under investigation. signals and temperature behaviour are
hot climates. The time constants of these For comparable standard crystalline comparable to those for c-Si. The oscilla-
effects are again temperature driven and only, the latest investigation of 24 c-Si tions between summer and winter can be
mainly related to FF. The average stabi- samples indicates a technological-origin- slightly lower for the samples with better
lised PSTC of the depicted a-Si 3 sample related difference of at least 5% (implying temperature behaviour or poor low-irradi-
is about –9% lower than that stated by again correct and stable nominal power ance behaviour. Any potential gains due
the manufacturer. Long-term degrada- values). This value increases greatly for to a better temperature coefficient can be
tion rates are superimposed onto these certain PV modules incorporating special lost again, however, as a result of higher
metastable effects. One module type technologies affecting energy yield perfor- average operating temperatures. The CIGS
completely failed the long-term test; two mance, such as in the case of bifacial PV 2 module shown in Figure 5 indicates
out of four samples ceased operation after modules or some thin-film technologies. the effect of an increasing PSTC over the

62 | May 2017 | www.pv-tech.org

 Design and Build Technical Briefing

years due to metastable behaviour, as Figure 6. Module

demonstrated earlier, which can be either performance
ratio MPRCalc,
positive or negative for PV modules of this
calculated using
type. weather data and
CdTe samples (green dots, Fig. 5) indoor measure-
show less oscillation by season, but still ments, plotted
versus the
exhibit maximum MPR values during the
measured MPROut-
winter months. The reasons for the lower door based on
amplitudes can be found in the signifi- energy-weighted
cantly lower temperature coefficient γ average outdoor
power (blue:
of typically –0.29%/K, and in the spectral
Cologne; green:
gains in summer. The difference between Ancona; red:
summer and winter is further reduced Tempe; orange:
because of the metastable behaviour, as Chennai)
shown earlier.
In the case of a-Si samples (red dots,
Fig. 5), the MPR values during the first
few months are dominated by Staebler–
Wronski degradation, followed by
temperature annealing observed in the
summer months. Compared with c-Si,
small oscillations between summer and
winter are achieved. In contrast to all other
cell technologies, the maximum MPR reach a stable level after more than a year values, can be achieved, as illustrated in
values are reached in summer; the reason of outdoor exposure. Figure 6; this deviation is assumed to be
is a combination of small temperature mainly due to the influence of PSTC measur-
losses, again due to low temperature Energy rating of PV modules using ing uncertainties on the MPROutdoor results.
coefficients γ (typically in the range of linear performance loss analysis The approach takes into account all the
–0.26%/K to –0.39%/K, depending on A linear performance loss analysis (LPLA), relevant factors that have an impact on
manufacturer), gains due to thermal as described in Schweiger et al. [9], can be energy yield, such as module temperature,
annealing, and significant spectral gains used to quickly, accurately and inexpen- low-irradiance conditions, and spectral
in summer. For some samples, high losses sively predict the MPR of PV modules and angular effects, as well as soiling.
due to poor low-irradiance behaviour in for different climates. Simple reference The approach also allows a quantifi-
winter were observed. It is noted that the environmental data sets and energy rating cation and comparison of the various
performance of some a-Si samples did not data, in accordance with the IEC 61853 influencing factors for different PV module
series, measured in the laboratory serve technologies and for different climates, as
as input data. An energy yield prediction illustrated in Figure 7. The energy yield of
Figure 7. Quantified loss mechanisms influencing the MPR of
PV module types c-Si 1, CdTe 1 and CIGS 2 in different climates based on calculated MPRCalc, with a devia- PV modules is affected by five individual
on an annual basis tion of ±3% from measured MPROutdoor loss factors; the mechanisms correspond

64 | May 2017 | www.pv-tech.org

 Technical Briefing Design and Build

to loss terms ∆MPR for different climates, spectral irradiance conditions is technol- Conclusions
which can be singled out. The loss mecha- ogy driven, but also depends on the Because of cost and time pressure,
nisms which influence the MPR of electri- individual manufacturers. The behaviour is consideration of the energy yield
cally stable PV modules are: temperature dominated by wafer recombination losses, performance of PV systems is often of
(∆MPRTEMP), low irradiance (∆MPRLIRR), and module internal serial and parallel secondary importance when construct-
spectral effects (∆MPRMMF), angular losses resistance in combination with operating ing PV plants. Optimisation of the yield
(∆MPRAOI) and soiling (∆MPRSOIL). voltage and current. The performance is necessary, however, for successful
The losses due to soiling and angular between different manufacturers may vary investment. Significant differences
effects are almost constant for PV modules significantly. A satisfactory low-irradiance were observed in the energy yield of PV
with standard untreated front glass. behaviour for constant spectral irradiance modules available on the market – up
Soiling losses (∆MPRSOIL) are highest in conditions means an efficiency drop of to 23%, depending on power rating,
Arizona, although higher soiling rates can less than –5% at 100W/m2 relative to STC; technology and climate.
be expected in Saudi Arabia. The soiling this can easily be tested in the laboratory. The results have shown that a combi-
rate is highly dependent on the period Losses due to temperature (∆MPRTEMP) nation of indoor tests and reference
under consideration, and long-term are highest for c-Si, with up to –9.6% in climate datasets is sufficient for estimat-
averages are needed. Chennai. Better values can be achieved ing, within ±3%, and comparing the
The losses due to angular effects with thin film when the advantages due to energy yield performance of different
(∆MPRAOI) are highest, compared with low temperature coefficients are not lost PV module technologies. The long-term
overall available energy, in Cologne, with because of higher operating tempera- stability of electrical power, however,
up to –3.5%. In addition to the advantages tures. must still be tested in the field.
gained in light transmission, lower angular The influence of spectral irradiance The ultimate owner of the PV power
losses can be achieved with deeply struc- (∆MPRMMF) on c-Si is low on an annual plant should consider a well-defined
tured glass (–2.8%) or an anti-reflection basis. The highest impact on energy yield module performance ratio before
coating (–1.6%). For deeply structured can be found for CdTe (up to +5.3% in making an investment decision. The
glass, however, higher soiling rates must Chennai) and a-Si. competitiveness of solar projects can
be considered. For other mounting conditions with be enhanced by PV modules with
Relative losses due to low-irradiance orientations that differ from optimal or reliable long-term performance and
behaviour (∆MPRLIRR) are also highest those with reduced ventilation, as in the optimal energy yield performance
in Cologne, with up to –3.6%. The case of building-integrated PV (BIPV), suited to the climate of the installa- 
low-irradiance behaviour for constant other loss factors must be assumed. tion location.

Acknowledgement Authors
This work was partially funded by the German Federal Markus Schweiger received the diploma in electrical engineering from
Ministry for Economic Affairs and Energy (BMWi) as part of the Technical University of Munich and is currently pursuing his doctorate
Contract Nos. 0325517B and 0325517C. degree in photovoltaics at the RWTH Aachen University. Since 2009 he has
been working as a scientist and project engineer in the Solar Innovation
Department at TÜV Rheinland. His expertise includes high-precision PV
module characterisation, performance behaviour of thin-film PV modules and energy
References yield of PV module technologies in different climates.

[1] UN Environment and Bloomberg 2017, “Global trends in renewable Werner Herrmann studied physics at the Technical University of Aachen and
energy investment” (Apr. 6). holds a PhD in PV power characterisation with solar simulators. Since 1988
[2] International Energy Agency (IEA) 2014, “Technology roadmap solar he has been working at TÜV Rheinland, and has gained more than 20 years’
photovoltaic energy”, 2014 edn. experience in developing standards for the PV industry. He is the team man-
[3] IEC 61853-1:2011, “Photovoltaic (PV) module performance ager for PV research activities that focus on PV module reliability testing,
testing and energy rating – Part 1: Irradiance and temperature output power characterisation, energy yield assessment, mechanical robust-
performance measurements and power rating”. ness of PV modules and PV safety.
[4] IEC 61853-2:2013, “Photovoltaic (PV) module performance testing
and energy rating – Part 2: Spectral response, incidence angle and Christos Monokroussos is a technical expert for TÜV Rheinland, where his
module operating temperature measurements” (IEC 82/774/CDV).
activities focus on R&D, characterisation of solar cells and PV modules,
[5] IEC 60891:2013, “Photovoltaic devices. Procedures for temperature quality control of measurement systems, standardisation progress and PV
and irradiance corrections to measured current voltage
characteristics”. module reliability. He earned his doctorate degree in photovoltaics from the
Centre for Renewable Energy Systems Technology (CREST) at Loughborough
[6] IEC 60904-7:2009, “Photovoltaic devices – Part 7: Computation
of the spectral mismatch correction for measurements of University, UK.
photovoltaic devices”.
[7] IEC 61215-1:2016, “Terrestrial photovoltaic (PV) modules – Design Uwe Rau studied physics at the University of Tübingen, Germany, and at
qualification and type approval”. Claude Bernard University Lyon, France. He received his diploma in 1987
[8] Schweiger, M. et al. 2017, “Performance stability of photovoltaic and his Ph.D. in 1991 in physics from the University of Tübingen, and also
modules in different climates, Prog. Photovolt: Res. Appl. earned his habilitation degree in experimental physics in 2002 from the
[manuscript under review]. University of Oldenburg, Germany. Since 2007 he has been the director of
[9] Schweiger, M. et al. 2017, “Understanding the energy yield the Institute for Energy and Climate Research at Forschungszentrum Jülich, and is a pro-
of photovoltaic modules in different climates by linear fessor at RWTH Aachen University, Germany. His research interests include the physics
performance loss analysis”, IET Renew. Power Gen. [DOI:10.1049/ and technology of thin-film solar cells, and the characterisation and qualification of PV
iet-rpg.2016.0682].
components.

www.pv-tech.org | May 2017 | 65