Monitoring 30,000 PV systems in Europe: Performance, Faults, and State of the Art

Abstract

We have tried to cast some light on some of the numerous questions concerning the performance of solar PV systems in Europe. We have based our analysis on the operational data monitored at more than 31,000 PV systems in Europe. These installations comprise residential and commercial rooftop PV systems distributed over 9 different countries, including multi-megawatt PV plants installed in the South of Europe. The PV systems were installed between 2006 and 2014. The mean Energy Yield of the PV systems located in the four reference countries are 1115 kWh/kWp for France, 898 kWh/kWp for the UK, 908 kWh/kWp for Belgium, 1450 kWh/kWp for the PV plants in Spain mounted on a static structure, and 2127 kWh/kWp for those mounted on a solar tracker in Spain. We suggest that the typical PR value for the PV systems installed in 2015 is 0.81. We have observed that the performance of the PV systems tends to increase when the peak power of the PV systems increases. We have found significant performance differences as a function of the inverter manufacturer, and the PV module manufacturer and technology. We have found an improvement of the state-of-the-art, in the form of an increase in performance in the yearly integrated PR of around 3 to 4% over the last seven years, which represents an increase of about 0.5% per year.

Full text

31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
1
Monitoring 30,000 PV systems in Europe: Performance, Faults, and State of the Art
Jonathan Leloux1,*, Jamie Taylor2, Rodrigo Moretón1, Luis Narvarte1, David Trebosc3, Adrien Desportes4
1Instituto de Energía Solar – Universidad Politécnica de Madrid (IES-UPM), Spain
2Sheffield Solar, University of Sheffield, UK
3BDPV, France
4Rtone, France
*Corresponding author: Jonathan Leloux (jonathan.leloux@ies-def.upm.es)
ABSTRACT: We have tried to cast some light on some of the numerous questions concerning the performance of
solar PV systems in Europe. We have based our analysis on the operational data monitored at more than 31,000 PV
systems in Europe. These installations comprise residential and commercial rooftop PV systems distributed over 9
different countries, including multi-megawatt PV plants installed in the South of Europe. The PV systems were
installed between 2006 and 2014. The mean Energy Yield of the PV systems located in the four reference countries
are 1115 kWh/kWp for France, 898 kWh/kWp for the UK, 908 kWh/kWp for Belgium, 1450 kWh/kWp for the PV
plants in Spain mounted on a static structure, and 2127 kWh/kWp for those mounted on a solar tracker in Spain. We
suggest that the typical PR value for the PV systems installed in 2015 is 0.81. We have observed that the performance
of the PV systems tends to increase when the peak power of the PV systems increases. We have found significant
performance differences as a function of the inverter manufacturer, and the PV module manufacturer and technology.
We have found an improvement of the state-of-the-art, in the form of an increase in performance in the yearly
integrated PR of around 3 to 4% over the last seven years, which represents an increase of about 0.5% per year.
Keywords: PV system, Performance Ratio, Energy yield, UK, France, Belgium, Spain, State of the art
1 INTRODUCTION
The number of solar photovoltaic (PV) systems
installed in Europe has drastically increased over the last
few years, mostly thanks to the advantageous feed-in
tariffs set in by each country’s government. A relatively
little fraction of the energy production data of these PV
systems has been analysed [1-6], and as a consequence,
there still remain wide gaps in the knowledge of the real-
world performance of these PV systems. This feedback
from the field is nevertheless important for the future
development of the PV industry and for the establishment
of new renewable energy development programmes by
the respective governments.
In this work, we have tried to cast some light on some
of the numerous remaining questions about the
performance of PV systems in Europe. To do so, we have
based our analysis on the operational data monitored at
more than 31,000 PV systems in Europe. These
installations comprise residential and commercial rooftop
PV systems distributed over 9 different countries, and
multi-megawatt PV plants installed in the South of
Europe. These PV systems were installed between 2006
and 2014, and the data have been measured at time
resolutions varying from minutely to monthly.
We have carried out an assessment of the energy
production and the performance of all of these
installations, from their commissioning date until the end
of 2014.
We have looked for important trends on these
installations, such as the decrease in performance with
the time of operation, or the increase in performance due
to improvements in the quality of PV system components
and installation practices. We have also studied the
distribution of the installed power per region, as a
function of the type of installation and the external
incentives.
We have characterised the state of the art of the PV
installations, both in terms of PV system components,
installation topology, orientation, and peak power,
differentiating from the small residential PV market in
northern and mid-latitude Europe and the PV plants in the
southern and sunnier countries. We have applied
statistical analyses on the whole database, which has
allowed us to analyse the main parameters that explain
the important differences that were observed among the
installations in terms of performance, such as the PV
module technology and manufacturer, the PV inverter,
the region and year of installation.
2 DATA AND METHODS
This work is based on the data collected at more than
31,000 PV systems in Europe (See Figure 1). Most of
them are located in the UK, Belgium and France. Spain
has few PV systems, but they are important ones because
most of them are large PV plants, which sum up a larger
total peak power than the rooftop installations from any
other country. About 300 PV systems are located in other
countries from Europe. This study is therefore mostly
based on the analysis of the data of 4 reference countries:
France, UK, Belgium (mostly small rooftop PV systems)
and Spain (mostly large PV plants). The diversity in
climates and installation types among these 4 countries
allows to reach a representative picture of the state of the
art in PV systems in Europe. The data on which this work
is based are publicly and freely available in a condensed
form from a web repository [7].
Figure 1: Location of the 31,000+ PV systems in our
database. Most of them are located in the UK, Belgium
and France.

31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
2
The geographical distribution and type of these data are
summarized in table I.
Table I: Geographical distribution and type of the PV
systems in our database. Most of the PV systems in Spain
are large PV plants. 300 PV systems are located in other
countries from Europe.
Country
# systems
Pp [MWp]
Type
France
17672
65
Rooftop
Belgium
7648
50
Rooftop
UK
5835
23
Rooftop
Spain
29
116
PV plant
Others
307
3
Rooftop
TOTAL
31491
255
Most of the data for the UK have been provided by
Sheffield Solar and were acquired through the Microgen
Database (MgDB) website [8]. Most of the data for
France and Belgium were provided by BDPV through its
free public website for PV system owners[9], and Rtone
through its commercial monitoring service Rbee Solar
[10]. The data from the PV plants in Spain have been
provided by several PV plant operators through Instituto
de Energía Solar - Universidad Politécnica de Madrid
(IES-UPM). The data for the rest of Europe have been
provided by Rtone and BDPV.
The data that were used as input for these analyses
can be divided into three main categories: PV systems
characteristics, PV energy production and solar
irradiation.
Some of the PV system characteristics were available
for most of the PV systems: latitude, longitude, azimuth
and tilt angles, and peak power. Some other
characteristics were available, depending on the data
provider: PV module manufacturer and model, inverter
manufacturer and model, installer, year of installation,
PV cell/module technology.
The PV energy production data provided by MgDB
were collected via the MgDB website, with PV owners
using the site as a portal to upload readings and in return
receiving free monthly Performance Ratio (PR) analysis
and peer-to-peer performance checking in the form of
interactive maps and nearest neighbour comparisons. The
majority of data is measured by the energy meters of the
inverters and is collected from commercial data donors
who own/monitor hundreds of systems using automated
data transfers. PR calculations interact directly with the
MgDB so as to provide regular updates to the live
website.
The PV energy production data provided by Rtone
have been monitored using the Rbee Solar monitoring
product, which measures the energy production with a
smart energy meter at a 10-min time interval. These
installations have been installed from 2008 to 2014.
The PV energy production data provided by BDPV
have been collected through a free, public website for PV
system owners, where individuals can manually enter the
monthly reading of their energy meter or inverter through
an HTML web interface, or if their equipment allows for
it, make use of an Application Program Interface (API) to
automatically upload their data, at any time resolution.
The PV energy production data provided by the PV
plant operators have been provided to IES-UPM in
several different formats, mostly in the form of text/csv
files. The oldest of these PV plants was commissioned in
2005, but most of them date from the years 2006-2010.
Figure 2 represents the relative distribution of the
azimuth and tilt angles of all the PV systems on a
heatmap. The figures are given in the format of a
percentage. Most of the installations have an azimuth
between South-East (-45º) and South-West (45º) and with
a tilt between 20º and 50º, which correspond to the
orientations for which the annual yearly yield is the
highest.
Figure 2: Relative distribution of the azimuth and tilt
angles of all the PV systems on a heatmap. The figures
are given as a percentage. Most of the installations have
an azimuth between South-East (-45º) and South-West
(45º) and with a tilt between 20º and 50º, which
correspond to the orientations for which the annual yearly
yield is the highest.
Figure 3 shows the frequency of the peak power of
the PV systems in Europe. Most of them have a peak
power between 1 and 6 kWp. A pronounced peak in
frequency is visible around 3 kWp. This peak is mostly
explained by more than 50% of installations in our
database coming from France, where the government has
set up a public incentive mechanism that makes it
uneconomical to install a PV system with a peak power
higher than 3 kWp. As a consequence, most of the PV
systems in France have a peak power of exactly 3 kWp or
slightly lower (see Figure 4).
Figure 3: Frequency of the peak power of the PV
systems in Europe. There is a pronounced peak around 3
kWp. This is a direct result of the public incentive
mechanism set up by the French government such that it
is uneconomical to install a PV system with a peak power
higher than 3 kWp. As a consequence, most of the PV
systems in France have a peak power of exactly 3kWp or
slightly lower. As a comparison, Figure 5 shows the
frequency of the peak power of the PV systems in
Belgium, where the public incentives have been
established following a very different mechanism. The
feed-in tariff set-up by the government establishes that
the owner of a PV system can sell as much PV-generated
electricity as the amount of electricity that he/she
consumes per year (annual net-balance mechanism). The
distribution of the peak power shows a wider dispersion,
with most of the PV systems between 1 kWp and 10
kWp, and it roughly follows a lognormal law, which

31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
3
arises as the product of several independent positive
random variables, the most relevant being the annual
energy consumption of the owner, the surface available
on the roof to install PV modules, and the capital that the
owner is willing/able to invest in a PV system. The feed-
in tariff in Belgium was so attractive that numerous
households have also found it economical to voluntarily
increase their annual electricity consumption in order to
be able to sell more PV electricity. They often did so by
switching to electricity for water heating, space heating,
or cooking.
Figure 4: Frequency of the peak power of the PV
systems in France. There is a pronounced peak around 3
kWp.
Figure 5: Frequency of the peak power of the PV
systems in Belgium. The distribution of the peak power
shows a wider dispersion, with most of the PV systems
between 1 kWp and 10 kWp, and it roughly follows a
lognormal law.
The monthly solar irradiation data have been
provided by the solar irradiation data provider Synermet
[11]. The monthly Global Horizontal Irradiation (GHI)
data were acquired from CM SAF [12], and they have
been translated into monthly Global Tilted Irradiation
(GTI) data using a decomposition and translation model.
The information reported by the data providers is not
always very accurate. We have observed these
inaccuracies in the information on some key parameters
such as the azimuth angle, the tilt angle, or the peak
power. We have also observed inaccuracies and
inconsistencies in the energy production data and in the
solar irradiation data. This leads to some important
uncertainties in our results, although we have tried to
minimise their impact, by applying filtering procedures,
and by using analyses methods that are robust against
erroneous data and outliers. As an example, Figure 6
shows the tilt angle information for all the PV systems in
our database as reported by the data providers. The
systematic peaks around values that are multiples of 5º
are an indicator that most of the data have been rounded.
Moreover, many tilt angles were reported as 0º, but they
do not appear on the figure, because they corresponded to
PV system orientations reported as 0º in tilt and 0º in
azimuth, which for some data providers is a convention
used as default when they do not know that information.
As very few PV generators are installed completely
horizontally, we have removed all of these data. It would
in any case be very bad practice to install the PV modules
horizontally, because they would tend to accumulate
soiling, and their energy yield would be greatly affected.
Figure 6: Tilt angle of all the PV systems in Europe as
reported by the data providers. The systematic peaks
around values that are multiples of 5º are an indicator that
most of the data have been rounded.
We have filtered out all the monthly integrated PR
data that were higher than 1.1, because these values are
very unlikely, and they are the consequence of
uncertainties on some of the input data. On the other
hand, we chose not to be too restrictive in the filterings in
order to avoid to skew our results by eliminating some
points that would look like outsiders but are due to other
causes that are physically sound. We have filtered out all
the yearly integrated PR data every time that at least one
of the months of this year for a given installation had
been filtered out.
We have applied a simple statistical filtering on the
yearly integrated PR. To do so, first we calculate the
median of the whole population (
)(
2/1 PRmedian

).
We then calculate the Median Absolute Deviation
(MAD) of the whole population, defined as:
 
2/1

 PRmedianMAD
and filter out any yearly
integrated PR value that is higher than
)5.1(3
2/1 MAD

and lower than
)5.1(6
2/1 MAD

. The application of this method
leads us to filter out all the PR values lower than 39.5%
and higher than 95.5%.
We have analysed the annual energy yield of the PV
systems, and we have calculated and represented its mean
value and dispersion for all the PV systems in Europe and
for four of the reference countries.
We have used the Performance Ratio (PR) to analyse
the overall performance of the PV systems.

31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
4
The PR is defined as:
r
f
Y
Y
PR
with final yield Yf defined by Epv/PSTC and the reference
yield Yr defined by GTI/GSTC where:
Epv is the net electrical energy produced by the PV
system during a given period of time, PSTC is the rated
power of the PV generator under Standard Test
Conditions (STC), GSTC is the global solar irradiance
under STC (i.e. 1000 W/m2), and GTI is the Global
Tilted Irradiation received by the PV generator.
We have analysed all the monthly integrated PR of
all the PV systems, and for all the months for which the
data were considered to be correct after filtering on a
month-by-month basis.
We have analysed the yearly integrated PR of all the
PV systems, and for all the years for which there were 12
consistent monthly data for the PV system.
We have used histograms and box plots to visualize
the data, and we have used an Analysis of Variance
(ANOVA) procedure to assess the statistical soundness of
our results for the comparisons of the PV systems’
performance as a function of several key parameters such
as the inverter manufacturer, the PV module
manufacturer, the PV module technology, the year of
installation, or the country where it is located. This
allows to test whether the differences observed between
different populations are significant enough relative to
the overall dispersion among the populations that is due
to uncertainties and/or the other parameters. We have
used multi-ANOVA procedures in order to analyze the
effects of one parameter on performance, while
normalizing as much as possible for the other relevant
effects. For example, when analyzing whether there is a
statistically significant relationship between a PV module
manufacturer and the performance of the PV systems
equipped with its PV modules, we have applied a multi-
ANOVA to verify that the results obtained regarding the
PV modules were not due to systematic associations
between one PV module and one inverter, where one
poorly performing inverter model or manufacturer could
negatively influence the results for the PV modules
associated with it.
3 RESULTS
3.1 Annual energy yield
Figure 7 shows the distribution of the annual energy
yield for Belgium, France, Spain, and the UK. The
distributions in Belgium and in the UK are very similar,
with a mean annual energy yield around 900 kWh/kWp.
This holds true in our database because a small
proportion of the systems from the UK are located in the
northern regions. The annual energy yield in France is
around 1,100 kWh/kWp, with a wide distribution that
largely reflects the difference in solar resource between
the north and the south of the country. The mean annual
energy yield in Spain appears to be around 1,900
kWh/kWp, but very few PV systems are encountered
around that value, which corresponds to a valley between
two peaks rather than to a peak. The explanation resides
in the existence of two types of PV plants: static and
trackers (See Figure 8). The mean annual energy yield for
static PV plants is around 1450 kWh/kWp, and the
corresponding value for PV plants with trackers is around
2,100 kWh/kWp. There is a wide variation in the annual
energy yield in Spain which, as with France, is largely
due to the difference in solar resource between the north
and the south of the country, but in the case of the PV
plants with trackers, these differences also reflect the
diversity in the tracking mechanisms.
Figure 7: Distribution of the annual energy yield for
Belgium, France, Spain, and the UK. The differences
observed are mainly due to the differences in the solar
resource available. In the case of the PV plants in Spain,
this also reflects the differences between static structures
and tracking mechanisms.
Figure 8: Distribution of the annual energy yield for PV
plants in Spain. The differences are largely due to the
difference in solar resource between the north and the
south of the country, but in the case of the PV plants with
trackers, these differences also reflect the diversity in
tracking mechanisms.
Table II summarizes the mean annual energy yield
for the 4 reference countries.
Table II: Mean annual energy yield data for the four
reference countries.
Country
Mean yield (kWh/kWp)
France
1115
UK
898
Belgium
908
Spain – PV plant – Static
1450
Spain – PV plant – Tracker
2127
3.2. Yearly integrated Performance Ratio
The yearly integrated PR has been calculated for all
the PV systems, and for all the years for which data were
available. Figure 9 shows the distribution of these yearly
integrated PR. The distribution does not follow a normal

31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
5
(or gaussian) distribution, because an important fraction
of the PV systems show an overall performance lower
than average, and others are clearly subject to faults,
which skews the distributions towards the low PR values.
The distribution is better explained with a Weibull
distribution, which often arises when the range of
variation of a population is physically limited at one
extremity, but not at the other. In the case of PV systems,
it is very difficult to reach yearly integrated PR higher
than 0.9, while it is much more likely that a PV system
will have a PR much lower than average, because of
performance problems.
Figure 9: The distribution does not follow a normal (or
gaussian) distribution, because it is skewed towards the
low PR values. The distribution is better fit by a Weibull
distribution.
We have tested the goodness of the Weibull
distribution on our data. To do so we have carried out a
probability plot, whose result is shown on Figure 10. The
probability plot shows that the Weibull explains very
well the PR distribution from PR values ranging from 0.6
to 0.9. This range of values represents the majority of the
PV systems. On the contrary, the Weibull distribution
does not explain the data for PR values lower than 0.6
and higher than 0.9, which is clearly visible from the
probability plot, and also from the P-value and from the
result of the Anderson-Darling (AD) test. For PR values
that are lower than 0.6, there is a significant departure of
the Weibull fit from the data, which shows that these data
belong to a different population. These PV systems are
subject to severe performance problems and to faults. On
the other hand, the Weibull law does not explain the PR
values that are higher than 0.9, which also suggests that
these PV systems belong to a different population. It is
very likely that these PR values are simply not real, and
that they are caused by uncertainties on solar irradiation
data, as well as on the azimuth and tilt angles and the
peak power of the PV systems. This probability plot
allows us to suggest that the yearly integrated PR values
ranging from 0.6 to 0.9 are representative of the state-of-
the-art for PV systems in Europe. We suggest to take the
scale (or peak or most frequent value) of the Weibull
distribution as the typical value that represents the whole
population. The shape of the Weibull fit to the data of all
the PV systems from Europe is around 0.79, and we
therefore suggest that this value is representative of the
typical PV system in Europe installed over the last few
years. Given the shape (the asymmetry) of this Weibull
distribution, the mean value of the PR is some 3% lower
than the most frequent value.
Figure 10: Probability plot of the yearly integrated PR
against a fit on the data using a Weibull law. The
probability plot shows that the Weibull law explains very
well the PR distribution from PR values ranging from 0.6
to 0.9. This range of values represents the majority of the
PV systems, and could represent the start of the art of the
PV systems installed in Europe over the last few years.
Figure 11 represents the distribution of the yearly PR
for the three reference countries whose installations are
predominantly rooftop-mounted: France, UK, and
Belgium. We take the typical (peak) values for yearly PR
are 0.81 in the UK, 0.80 in Belgium, and 0.78 in France.
Figure 11: Distribution of the yearly PR for the three
reference countries whose typology corresponds to
rooftop installations: France, UK, and Belgium. The
typical (peak) values for yearly PR are 0.81 in the UK,
0.80 in Belgium, and 0.78 in France.
Figure 12 shows the distribution of the yearly
integrated PR for PV plants in Spain (both with static
structures and trackers). The typical value of PR is 0.81.
PV systems in Spain are more affected by thermal losses
in the PV cells than the PV systems in the other countries
of Europe. Notwithstanding, the PR obtained for the PV
plants in Spain is still high when compared to the other
countries in Europe. This high value for the PV plants in
Spain relatively to the other countries reflects the higher
efficiencies of larger PV installations, that take benefit
from scale effects.

31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
6
Figure 12: Distribution of the yearly PR for PV plants in
Spain. The typical value is 0.81.
Figure 13 shows the comparison of the yearly
integrated PR of the 4 reference countries on a boxplot.
The middle line of the boxes represents the median, or
the 50th percentile of the distribution. The bottom line
represents the 25th percentile, whilst the upper line
represents the 75th percentile. The whiskers indicate the
dispersion among the rest of the population, excluding
outliers. The PV plants in Spain and the rooftop
installations in the UK appear to have the higher median
value, followed by Belgium, and then France. The PR in
France is some 3% lower than in the UK, and 2% lower
than in Belgium.
Figure 13: Comparison of the yearly integrated PR of the
4 reference countries using boxplots. The PV plants in
Spain and the rooftop installations in the UK appear to
have the highest median value, followed by Belgium, and
then France. The PR in France is some 3% lower than in
the UK, and 2% lower than in Belgium.
Table III summarizes the result of the yearly PR
calculations for the 4 reference countries.
Table III: Yearly PR for the four reference countries.
Country
Mean PR [-]
Typical [-]
France - Rooftop
0.75
0.78
UK - Rooftop
0.78
0.81
Belgium - Rooftop
0.77
0.80
Spain - PV plant
0.78
0.81
3.3. Influence of the peak power on performance
Figure 14 shows the relationship between the yearly
integrated PR and the peak power of the PV system. We
observe an increase in performance along with the
increase in the peak power of the PV system. This is in
particular true for PV systems whose peak power is lower
than 4 kWp. This is mostly explained by the increasing
yield of the inverter as a function of their nominal power,
but in some cases, in particular for large PV installations,
the higher performance also arises from more demanding
quality controls. This advocates for more quality control
procedures in small rooftop installations. This also
suggests that when the policymaker decides to shape the
PV landscape towards small systems, such as it was the
case in France with systems limited to 3 kWp, this is
done at the price of lower PV system performance. An
ANOVA allowed us to confirm that the trend observed is
independent of external factors.
Figure 14: Relationship between the yearly integrated
PR and the peak power of the PV system. We observe an
increase in performance along with the increase in the
peak power of the PV system. This is in particular true
for PV systems whose peak power is lower than 4 kWp.
3.4. Influence of the inverter on performance
We have attempted to assess whether there was a
significant difference in performance between PV
systems equipped with inverters produced by different
manufacturers. To do so, we have grouped the PV
systems by inverter manufacturer, and we have calculated
the yearly integrated PR for each one of these sub-
groups. Figure 15 presents the result of this exercise, for
the inverters that equip at least 100 PV systems in our
database.
Figure 15: Boxplot representing the yearly PR of the
inverters present on at least 100 PV systems in our
database. We observe that there is a difference in the
median value of PR of some 5% between the best
performer and the worst performer.
We observe that there is a difference in the median
value of PR of some 5% between the best performer and
the worst performer. This seems to be partly in line with
the datasheets of the inverter manufacturers, some of

31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
7
which deliver higher energy yields than others, but it also
suggests that under real conditions, the real performance
of some inverters depart more from their nominal
performance than others. An ANOVA allowed to confirm
that the differences observed between the best and the
worst inverter manufacturers had enough statistical
power.
3.5. Influence of the PV module on performance
We have also attempted to assess whether there was a
significant difference in performance between PV
systems equipped with PV modules produced by different
manufacturers (except for the thin-film technologies, that
are dealt with later on). Figure 16 presents the result of
this exercise, for the PV modules that equip at least 100
PV systems in our database. The names of the
manufacturers have been masked under a hidden code.
The country of the manufacturers corresponds to the
headquarters of the company, and not necessarily to the
location of the manufacturing, which in some cases may
be different. We observe that there is a difference in the
median value of PR of some 6% between the best
performer and the worst performer. This seems to be
partly in line with the datasheets of the PV module
manufacturers, most of which have a nominal power
tolerance of +- 3%. Therefore, these differences could
arise if some manufacturers sell their PV modules with a
tolerance that is very close to + 3%, while some others
sell with tolerances very close to - 3%. An ANOVA
allowed to confirm that the differences observed between
the best and the worst PV module manufacturers had
enough statistical power. From the figure, we can also
observe that there is no clear correlation between the
geographical origin of a PV module manufacturer and the
quality of the PV module.
Figure 16: Boxplot representing the yearly PR for the PV
modules that equip at least 100 PV systems in our
database (excluding thin-film technologies). We observe
that there is a difference in the median value of PR of
some 6% between the best performer and the worst
performer.
We have also attempted to study the impact of the PV
cell/module technology on the performance of the PV
systems. The results are shown on Figure 17. The classic
and most common crystalline silicon technology, marked
as xSi, can serve as a reference to assess the performance
of the other technologies relatively to it. We observe that
two technologies perform slightly better than the xSi
technology. These technologies are the Heterojunction
with Intrinsic Thin layer (HIT), and the back-contact
silicon (bcSi). Other technologies show performances
that are lower than the ones of the xSi. Immediately after
the xSi technology comes the Saturn silicon (xSi-Sat).
Then comes the Cadmium Telluride (CdTe). The
dispersion of its performance appears to be higher, but
this is mostly because the information that we possess on
this technology comes from PV plants in Spain rather
than in rooftops in Europe. The sample therefore appears
as composed of less elements in the statistical analysis,
although the total peak power represents around 30 MW.
Furthermore, the dispersion mostly arises from one single
PV plant that has been subject to severe performance
problems for several years, related with the PV modules
themselves (this PV plant appears at the lowest PR in
Figure 12, with a value of 0.55). Nevertheless, as it can
be observed, the median value of the entire technology is
not much lower than for the xSi-Sat technology, and it is
probably more representative of the newer modules made
using this technology. Then comes the Upgraded
Metallurgical Grade Silicon (UMGSi). The last two
boxplots are for thin-film technologies and show a
performance that is considerably lower than the rest of
the technologies. The second poorest performer is the
amorphous silicon (aSi). The worst performer is the
Copper Indium Selenide (CIS) and the Copper Indium
Gallium (di)selenide (CIGS). Nevertheless, in this last
case, the information that we had at our disposal only
came from two PV module manufacturers, and both have
now ceased their activities in PV, so our result should not
be taken as representative of the latest CIS/CIGS
technologies. An ANOVA allowed to confirm that the
differences observed as a function of the PV module
technology had enough statistical power.
Figure 17: Boxplot representing the yearly PR as a
function of the PV cell/module technology. HIT =
Intrinsic Thin layer; bcSi = back-contact silicon; xSi =
classic crystalline silicon; xSi-Sat = Saturn silicon; CdTe
= Cadmium Telluride. UMGSi = Upgraded Metallurgical
Grade Silicon; aSi = amorphous silicon; CIS/CIGS =
Copper Indium Selenide (CIS)/Copper Indium Gallium
(di)selenide (CIGS).
3.6. Evolution of the state of the art
We have investigated whether it was possible to
observe any improvement in the state of the art of PV
systems over time that would directly translate into an
increase in overall performance. To do so, we have
grouped the PV systems as a function of their year of
installation. We have compared the yearly integrated PR
of these groups during 2013 as a reference year. The
result is shown on Figure 18, where we observe a
remarkable increase in systems’ performance over time,
with an increase in yearly PR around 3 to 4% over the
last seven years, which represents an increase of about
0.5% per year. For this result, we can also suggest that

31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
8
the state of the art of PV systems in Europe installed
today is probably some 2% higher than the mean results
that we have obtained with our data ranging from 2007 to
2014. Therefore, because the typical yearly PR for that
period was assessed to be around 0.79, we can probably
assert that the state of the art for PV systems installed in
2015 corresponds to a typical yearly PR around 0.81.
Figure 18: Yearly PR in 2013 for all the PV systems,
grouped as a function of their year of installation. We
observe an increase in systems’ performance over time,
with an increase in yearly PR around 3-4% over the last
seven years, which represents an increase of about 0.5%
per year.
3.7. Degradation of PV system performance
We have attempted to assess whether it is possible to
observe any relevant degradation of the PV systems’
degradation over time. We have done so by analysing the
yearly integrated PR of all the PV systems installed in the
year 2008, as a function of the year of production from
2009 to 2014. We chose the systems installed in 2008
because it is the first year for which we have at least 1000
PV systems with yearly energy production data that that
have passed the filterings. The result is shown on Figure
19, where we can observe that the degradation in PV
systems’ performance over the last six years ranges from
unnoticeable to very low. The decrease in performance
from 2009 to 2014 appears to be around 1.5%.
Nevertheless, an ANOVA that was carried out on these
results showed that it was not possible to conclude with a
high statistical power that this observed decrease was not
merely the effects of the uncertainty, or reversely that the
mean degradation is accurate. Therefore we have not
been able to quantify the degradation.
Figure 19: Yearly integrated PR of all the PV systems
installed in the year 2008, as a function of the year of
production from 2009 to 2014. The degradation in PV
systems’ performance over the last six years ranges is
very low, and we have not been able to quantify it.
4 DISCUSSION
Apart from our analyses carried out on yearly
integrated data, whose objective was to reach a global
understanding of the performance of the PV systems in
Europe, we have analysed the distribution of the monthly
integrated PR for all the PV systems in Europe to
investigate whether other complementary observations
could be obtained from them. The result is shown on
Figure 20. The dispersion in the monthly integrated PR is
much higher than that observed for the yearly integrated
PR for several reasons. Firstly the uncertainty on PR is
higher during the winter months, which is made more
explicit when monthly data are analysed independently.
Secondly, any performance problems that span one
month or less are much more visible on the monthly data
than on the yearly data. This dispersion is therefore a
strong indicator of the presence of performance problems
or faults on some of the PV systems. More evidence of
short-lived faults appears when we move towards higher
temporal resolutions e.g. hourly data. The description of
these kinds of performance problems and faults are out of
the scope of this work, but the evidence of their presence
suggest that they should be studied more carefully in
future works.
Figure 20: Dispersion in the monthly integrated PR. It is
much higher than the one observed for the yearly
integrated PR.
Even though we have analysed more than 30,000 PV
systems in Europe, these were mostly located in four
countries. We have analysed many PV module and
inverter manufacturers and technologies, but not all of
them. Furthermore, new technologies have recently
appeared in the market, and it would be interesting to
analyse them in order to update our results. The
uncertainties around many of our analyses are important,
and further development may enable us to reduce them in
order to obtain more accurate and more in-depth
conclusions. We still have many gaps to cover to achieve
a comprehensive understanding of the state-of-the-art of
PV systems in Europe. With this in mind, we welcome
any future collaboration that might improve our outlook,
whether by completing our database with other countries
and technologies, or by comparing methods and results.
By definition, a review of the state-of-the-art of PV
systems requires regular iterations in order to continue to
incorporate the latest advances.

31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
9
5 CONCLUSION
We have analysed the data from more than 31,000 PV
installations in Europe. The mean Energy Yield of the PV
systems located in each of the four reference countries is
1115 kWh/kWp for France, 898 kWh/kWp for the UK,
908 kWh/kWp for Belgium, 1450 kWh/kWp for the PV
plants in Spain mounted on a static structure, and 2127
kWh/kWp for those mounted on a solar tracker in Spain.
We have shown that the distribution of the yearly
integrated PR can be modeled well using a Weibull
distribution for PR values ranging from 0.6 to 0.9. This
range of values represents the majority of the PV
systems, and we suggest that they are representative of
the state-of-the-art for PV systems in Europe. We suggest
that the typical PR value for the PV systems installed in
Europe over the last few years is 0.79, and that the value
for the PV systems installed in 2015 is 0.81. The
corresponding mean values are respectively 0.76 and
0.78.
We have observed that the performance of the PV
systems tends to increase when the peak power of the PV
systems increases, in particular for PV systems whose
peak power is lower than 4 kWp.
We have grouped the PV systems as by inverter
manufacturer and we found significant performance
differences, up to 5%, between the best performer and the
worst.
Similarly, we have grouped the PV systems as a function
of the PV module manufacturer, and have found
significant performance differences, up to 6%, between
the best performer and the worst.
We have found significant differences in performance as
a function of the PV module/cell technology.
We have found an increase in performance in the yearly
integrated PR around 3 to 4% over the last seven years,
which represents an increase of about 0.5% per year.
We have qualitatively observed some performance
degradation of the PV systems over the last few years but
we were not able to reliably quantify it without further
investigation.
We still have many gaps to cover to reach a
comprehensive and accurate understanding of the state-
of-the-art of PV systems in Europe. We therefore
welcome any future collaboration towards this goal.
ACKNOWLEDGMENTS
The Microgen Database is a public-industry-academic
collaboration providing solar photovoltaic performance
data for use across the UK PV supply chain. The
Sheffield Solar project is funded by the EPSRC (Solar
Energy for Future Societies: EP/I032541/1; Wise PV:
EP/K022229/1) and The University of Sheffield.
Thanks to Maria-Madalina Opincaru and Sam Chase for
their roles as database admin for the MgDB.
The work of Jonathan Leloux has been partially financed
by the European Commission within the project
PVCROPS [13] under the 7th Framework Program (Grant
Agreement nº 308468).
We feel indebted towards the thousands of PV system
owners and the PV plant operators who have freely
donated their data through BDPV, through MgDB, or
directly to IES-UPM.
REFERENCES
[1] G. Colantuono, A. Everard, L. M. Hall and A. R.
Buckley, Monitoring nationwide ensembles of PV
generators: Limitations and uncertainties. The case of the
UK, Solar Energy, vol. 108, pp. 252-263, 2014.
[2] J. Leloux, L. Navarte and D. Trebosc, “Review of the
performance of residential PV systems in Belgium,”
Renewable and Sustainable Energy Reviews, vol. 16, no.
1, pp. 178-184, 2012.
[3] J. Leloux, L. Navarte and D. Trebosc, “Review of the
performance of residential PV systems in France,”
Renewable and Sustainable Energy Reviews, vol. 16, no.
2, pp. 1369-1376, 2012.
[4] N. H. Reich et al., Performance ratio revisited: is
PR > 90% realistic?, Progress in Photovoltaics, 20: 717–
726, 2012.
[5] A. Woyte et al., Monitoring of photovoltaic systems:
good practices and systematic analyses, 29th European
Photovoltaic Solar Energy Conference and Exhibition,
2014.
[6] U. Jahn and W. Nasse, Operational performance of
grid-connected PV systems on buildings in Germany.
Progress in Photovoltaics: Research and Applications,
2004.
[7] Jonathan Leloux, Jamie Taylor, Rodrigo Moretón,
Luis Narvarte, David Trebosc, Adrien Desportes, PV
systems performance in Europe: dataset from 30,000 PV
systems, Public Dataset, 2015
(https://www.researchgate.net/publication/283210790_P
V_systems_performance_in_Europe_dataset_from_3000
0_PV_systems).
[8] Sheffield Solar, Microgen Database (MgDB],
Sheffield Solar - University of Sheffield
(http://www.microgen-database.org.uk/).
[9] BDPV: Base de Données PhotoVoltaïque
(http://www.bdpv.fr/).
[10] Rtone Rbee Solar monitoring solutions
(http://www.rtone.fr/).
[11] Synermet (http://www.synermet.com/).
[12] CM SAF (http://www.cmsaf.eu/).
[13] PVCROPS, European Commission FP7 project
(http://www.pvcrops.eu/).