Content uploaded by Jonathan Leloux
All content in this area was uploaded by Jonathan Leloux on Oct 11, 2017
Content may be subject to copyright.
SOLAR ENERGY INSTITUTE
POLYTECHNIC UNIVERSITY OF MADRID
Review of the
Residential PV Systems
Jonathan Leloux1,* Luis Narvarte1, David Trebosc2
1 Instituto de Energía Solar – Universidad Politécnica de Madrid, Photovoltaic Systems Group, EUITT,
Carretera de Valencia, km 7, 28031, Madrid, Spain.
2 BDPV, Rue du Capitaine Fracasse, 1, 31320, Castanet Tolosan, France.
Scientific journal article published in Renewable and Sustainable Energy Reviews, Elsevier, 2012
Jonathan Leloux, Luis Narvarte, David Trebosc, Review of the performance of residential PV systems in France,
Renewable and Sustainable Energy Reviews, Volume 16, Issue 2, February 2012, Pages 1369-1376, ISSN 1364-
The main objective of this paper is to review the state of the art of residential PV systems in France. This is
done analyzing the operational data of 6868 installations. Three main questions are posed. How much
energy do they produce? What level of performance is associated to their production? Which are the key
parameters that most influence their quality? During the year 2010, the PV systems in France have
produced a mean annual energy of 1163 kWh/kWp. As a whole, the orientation of PV generators causes
energy productions to be some 7% inferior to optimally oriented PV systems. The mean performance ratio
is 76% and the mean performance index is 85%. That is to say, the energy produced by a typical PV
system in France is 15% inferior to the energy produced by a very high quality PV system. On average, the
real power of the PV modules falls 4.9% below its corresponding nominal power announced on the
manufacturer’s datasheet. A brief analysis by PV modules technology has led to relevant observations
about two technologies in particular. On the one hand, the PV systems equipped with heterojunction with
intrinsic thin layer (HIT) modules show performances higher than average. On the other hand, the systems
equipped with the copper indium (di)selenide (CIS) modules show a real power that is 16 % lower than their
Keywords: residential PV system, energy production, performance ratio, performance index, CIS, HIT, PI,
The main objective of this paper is to review the state of the art of residential PV systems in France. This is
done analyzing the operational data of a representative sample of 6868 installations, totalizing a peak
power of approximately 20.5 MW, and installed between 2007 and 2010. At the end of March 2011, 1146
MW were installed in Metropolitan France. About half of the total power is installed in residential PV
systems of less than 10 kWp. The database here considered represents approximately 3.5% of the
residential PV in Metropolitan France at the end of March 2011.
The study articulates the analysis around three questions:
1) How much electricity do PV systems produce in terms of kWh per kWp?
2) What is their performance for producing electricity? The PV systems quality is analyzed using
different performance indicators such as the performance ratio (PR), the performance ratio at
standard test conditions (STC), condensed as PRSTC and the performance index (PI).
3) Which are the key aspects that influence the quality of PV systems? Statistical tools are applied to
find them out.
For the first question, related to energy production, a survey is realized over the monthly energy production
data supplied by the PV systems’ owners through a Website.
For the second question, related to the performance of PV installations, the assessment is based on the
aforementioned performance indicators, all of them consisting on comparing the real energy production of
each of the systems with the production simulated for a corresponding hypothetical system used as a
For the third question, an Analysis-of-Variance (ANOVA) applied on the PI uncovers the key aspects that
influence the quality of PV systems. A general multidimensional ANOVA is realized by grouping the PV
systems according to four characteristics: PV modules manufacturer, inverters manufacturer, installer, and
PV generator power. The goal is to isolate the causes explaining the PI differences.
The results presented in this work allow extracting conclusions about the expected energy production of PV
residential systems representative of the state of the art. They quantify the energy production losses due to
the orientation of the PV generators. The important quantity of PV systems analyzed makes it possible to
extend the results not only to the French market, but also to the European one and, hence, they are of
general interest. In fact, the conclusions are congruent with previous analyses of the operational
performance of residential PV systems installed during the last two decades in Germany, Switzerland, Italy,
Spain, Netherlands, Japan and USA[3-6], and can be useful to important works that are presently ongoing
and whose main purpose is the assessment of the performance and reliability of PV systems.
2. French residential PV market
The data analyzed in the present study concerns Metropolitan France (i.e excluding Overseas France). PV
experienced there an important growth since the year 2004, with the establishment of a tax credit of 40% of
the PV system cost. The growth was accelerated in 2005 with the rise of that tax credit up to 50%. But the
decisive moment was the vote in 2006 of a new feed-in tariff specific to PV of at least 0.30 €/kWh, and up
to 0.46 €/kWh for Building Integrated Photovoltaics (BIPV). As a direct consequence, from the year 2007,
the number of residential PV systems started to take off, reaching 20 MW at the end of that year. At the end
of March 2011, residential PV systems represented more than 550 MW. That power was distributed over
more than 160,000 installations.
Fig. 1 shows that 98% of the residential PV systems installed in France have a peak power of 3 kWp or
less, and more than half of the installations have a peak power very close to 3 kWp. This situation arose as
a direct consequence of a legal frame that strongly discourages installations of more than 3 kWp, mainly for
two reasons. First, the tax credits are denied for the PV systems of more than 3 kWp. Second, a VAT of
5.5% is applied to systems of less than 3 kWp, while it jumps to 19.6% for systems of more than 3 kWp. The
power distribution among residential PV systems in France is thus mainly explained by legal
considerations, rather that technical ones.
Fig. 1. Histogram of the peak power of PV systems. Nearly 98% of the residential PV systems installed in
France have a peak power of 3 kWp or less, and more than half of the installations have a peak power very
close to 3 kWp.
In our database, PV modules based on classical crystalline silicon (xSi) technology represent about 78% of
the total market shares. The rest of the market is distributed, by order of importance, among heterojunction
with intrinsic thin layer (HIT) (17%), amorphous silicon (aSi) (2%), copper indium (di)selenide (CIS) (2%),
and cadmium telluride (CdTe) (1%).
The database contains 106 PV modules manufacturers, 21 inverters manufacturers and 423 PV systems
installers. Fig. 2 shows that the relative market penetration within PV modules and inverters manufacturers
is satisfactorily modeled by a power-law, indicating that the market is dominated by a reduced number of
actors. The most extreme case of market domination is the control of more than 50% of the market by one
single inverter manufacturer (the point corresponding to that inverter manufacturer is not represented in
Fig. 2 because it is out of scale, but is taken into account in the power-law equation). The leading PV
modules manufacturer distributes a HIT technology and has a market share of 17%. The installers market
does not follow a power-law mainly because many installers are small familiar enterprises that only work at
local scales, much smaller than the country.
Fig. 2. Relative portion of the PV industry outstripping a given market share. The relative market
penetration for PV modules and inverters manufacturers is satisfactorily modeled by a power-law. The R-
squares of the fits yield 97.5% for the PV modules, and 91.9% for the inverters.
3. Performance assessment methodology
3.1. Energy production
As mentioned earlier in the text, the data concerning the PV systems were supplied by their owners. Each
PV system is localized by its latitude and longitude, completed with the corresponding altitude. The PV
generator is characterized by its orientation and tilt angles, its total surface, and its total peak power. The
data also provides information about the manufacturers of the PV modules and inverters that equip the
system, and the installer. The net energy production is reported on a monthly basis, and is read at the
inverter (95% of the database), or at the meter (20%), or at both sources (15%). The PV owners also
communicated the annual energy that they expected to produce, and that was generally estimated by the
installer before the commissioning of the installation. Not all the PV owners reported the energy production
corresponding to each month, and only 25% of them reported it systematically and correctly.
Thanks to the PV owners that simultaneously provided the energy production data coming from both the
inverter and the meter, it was possible to compare both sources of information. Fig. 3 shows the results of
these comparisons. The ratio Einverter/Emeter shows values ranging from 0.93 to 1.09. A ratio superior/inferior
to 1 indicates that the inverter systematically overestimates/underestimates the energy produced. When the
inverters are grouped by manufacturer and model, these ratios show a much lower dispersion, which
shows that some inverters systematically overestimate/underestimate the energy that is really produced.
Therefore, in the present study, the data provided by the inverters are adjusted by comparison with the data
provided by the energy meters.
Fig. 3. Histogram of the relation between the energy metering by inverter and meter
3.2. PV systems performance
The performance of PV systems is realized by comparison with a corresponding reference system. The
simulations require the input of the horizontal solar radiation and the ambient temperature data, both on a
monthly basis, which have been obtained from SoDa and PVGIS respectively. The solar radiation
received on the surface of each of the PV generators is estimated using widely accepted solar radiation
models[10-12]. The estimation of the energy production of the reference system is simulated with a tool
developed at IES-UPM and based on widely accepted models, whose details have been described
The energy performance indicators that are used to assess the technical quality of a particular PV system
are obtained by comparing its actual production along a certain period of time with the production of a
hypothetical reference system (of the same nominal power, installed at the same location, and oriented the
same way) somewhat free of certain kinds of losses. Table 1 presents three different performance
indicators that are used to assess the quality of a PV system. All three compare the real energy production
of the PV system during a certain period of time to the corresponding reference system. The variation
between them comes from the different reference system that is chosen in each case.
The PR is, by far, the most widely used performance indicator today, because the unitary energy
production, which is of paramount importance for economic analyses, is simply given by the product of the
irradiance, (or the number of “sun-hours”) by the PR. The difference between 1 and PR lumps together all
imaginable energy losses (real power of the PV modules power below nominal rating, mismatch, wiring,
shades, dust, thermal, DC/AC, failures, etc.). Because thermal losses are site-dependent (they depend on
climate), the PR of a given, unchanged PV system fluctuates from one place to another, and along the
course of a year or a day, which represents an obvious inconvenient for strictly qualifying its technical
quality. The PR of a given PV system located in the North of France is higher than the PR of the same
system installed in the South of the country. The PRSTC takes away such thermal losses, which requires to
consider (measure or estimate) the temperature of operation of the solar cells. Because of that, it is of more
complex calculation than the PR, but it becomes practically independent from time and site, thus being
more appropriate for strictly qualifying technical quality on a comparative basis. However, the PRSTC value
corresponding to an excellent quality and properly maintained PV system is lower than 1, mainly because
real inverters always associate some energy losses to the DC/AC conversion. Hence, one further step can
still be taken subtracting the DC/AC conversion losses corresponding to a top class inverter, let us say, one
whose European efficiency is 96%. That leads to the so called PI. It should be noted that a PI =1
corresponds to a PV system composed by an inverter and a PV generator whose real power and
characteristics coincide with their rated nominal value, free of shading, dust and wiring losses and also free
of failures. Consequently, the difference between 1 and PI can be understood as a measure of the
somewhat avoidable energy losses. The PI thus allows comparing directly the quality of PV systems under
different climatic and installation conditions. Because of that, this paper pays particular attention to the
analysis of PI values. Fig. 4 shows the evolution during the year 2010 of both PI and PR for a typical PV
system of the sample, free of shading, not experiencing any lack of availability or other second order
problems, whose PI is 84%, whose PRSTC is 80.5% and whose PR is 76.5% (all in annual values). The PI is
relatively constant along the year, while the PR varies of some 10% between winter and summer, mainly
due to the evolution of cell’s temperature. This lesser fluctuation of PI respect to PR suggests that PI is a
better quality indicator of the quality of PV systems than PR.
Table 1: The quality of a PV system for producing energy can be described through three different
performance indicators: PR, PRSTC or PI.
where Eproduced 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 STC, GSTC is the global solar
irradiance under STC (i.e. 1000 W/m2),
and G is the global solar irradiance
received by the PV generator.
Free of any kind of
system losses. Its solar
cells are always kept at
where PSTC represents the thermal
power losses in the PV cells due to
their operational temperature which is
different than STC.
Free of any kind of
system losses. Its solar
cells operate at the
same temperature that
the ones of the system
to be compared with.
where PDC/AC represents the
conversion losses due to the inverter
that equips the reference system.
High quality PV system.
Almost free of system
losses, except mainly
the cell’s temperature
losses and the DC/AC
losses corresponding to
a very good inverter and
Fig. 4. Evolution of PI, PRSTC and PR for a PV system during the year 2010.
Table 2 summarizes step by step the methodology used to calculate the performance indicators.
Table 2: General methodology used for the assessment of the performance of residential PV systems.
I. Data Collection at each location
- PV systems monthly real energy production 
- PV systems main characteristics: PV generator peak power, surface, tilt and
orientation, PV modules and inverters models, installers, general comments about
the system 
- Monthly global horizontal radiation 
- Monthly Tmax and Tmin 
II. Solar Radiation on PV generators
- Clearness indexes for global and diffuse radiation 
- Daily global, direct and diffuse radiation 
- Global radiation on PV generator surface 
III. Calculation of performance ratio (PR)
- Rated power under STC 
- IV curve under outdoor conditions 
- PR = Ratio (Real Energy Production / Energy Production without system losses)
IV. Calculation of performance index (PI)
- Losses due to cell temperature 
- Spectral losses [16, 17]
- Inverter electrical model 
- PI = Ratio (Real Energy Production / Energy Production for Reference System)
3.3. Statistical analysis on the parameters affecting the performance of PV systems
To investigate furthermore the main causes of the quality differences observed among the PV systems,
they have been grouped by common properties. The statistical method Analysis-of-Variance (ANOVA) has
been used to study the causes of the dispersion of PI. ANOVA procedures rely on a distribution called the
F-distribution. The key statistic is F = MSTR/MSE, where MSTR (Mean Square Treatment) represents the
variation among the means of the different groups, and MSE (Mean Square Error) represents the variation
within the groups. Large values of F indicate that the variation among the groups is large relative to the
variation within the groups, and hence that the groups are significantly different. A general multidimensional
ANOVA was realized according to four grouping criteria: PV modules manufacturer, inverters manufacturer,
installer, and PV system peak-power.
4.1. Energy production
The energy production analysis is carried out for the year 2010 and for the 1635 PV systems from which
the monthly production was correctly reported for the 12 months of the year. Fig. 5 shows a histogram of
those energy productions. On average, the PV systems produced in 2010 a net annual energy of 1163
kWh/kWp. The dispersion is mainly due to three factors: geography (and therefore solar radiation and
temperature), orientation and performance. The solar radiation during the year 2010 in France was globally
comparable to the mean radiation during the last decade. The energy productions reported are thus
sufficiently representative to be compared with other previous studies in the literature. As a comparison,
annual productions around 800 kWh/kWp were reported for PV systems installed 5-10 years ago in the
North and East of Germany. Two main causes explain the lower productions reported for the PV systems
in Germany respect to France. First, the solar radiation is globally higher in France. Second, the energy
productions reported in Germany correspond to PV systems installed about 10 years ago, whose quality
was probably lower, and whose power has decreased with time, mainly due to the light soaking.
Fig. 5. Histogram of the production of the PV systems in 2010.
4.2. Energy losses related to tilt and orientation
The vast majority of PV generators have a tilt angle between 20 and 50 degrees, which generally
corresponds to the configuration of the roofs on which they are mounted. At the latitudes of France, from
43º to 51º North, a PV generator maximizes its annual energy produced when it faces South and benefits
from a tilt angle around 40º. When the orientation is different, which is usual in residential PV, the energy
produced diminishes by an amount that is shown in Fig. 6. That same figure also shows the relative
distribution, in percent, of the number of residential PV systems installed, in function of the orientation and
tilt. It is worth underlying that low tilt values favor dust accumulation (tilt angles of less than 10º have been
reported to keep hold of important quantities of dust), but Fig. 6 shows that it is not frequent to find those
low tilt values.
Fig. 6. Distribution of the number of PV systems installed (out of a total number of 6868), in function of the
orientation and tilt, together with the corresponding net annual energy produced by a PV system in France
respect to the optimal inclination, in percent.
Fig. 7 shows the relation between the energy losses due to orientation and the proportion of PV systems
installed. It is satisfactorily described by a power-law (R2 = 97.5%). Almost 65% of the PV systems loose
less than 5% of their annual energy due to orientation, and less than 10% lose more than 15%. As a whole,
the orientation of residential PV causes energy productions to be 7% inferior to optimally oriented PV
systems, which can be interpreted as the price to pay, in terms of energy losses, for installing PV systems
on roofs instead of installing PV farms.
Fig. 7. Proportion of PV systems (in %) oriented so that they lose less than a given percentage of net
annual energy respect to the optimum orientation. Almost 65% of the PV systems lose less than 5% of
annual energy due to their orientation, and less than 10% lose more than 15%.
4.3. Performance of PV systems
Fig. 8 presents the histogram of PR and PI of 1635 PV systems that correctly provided the monthly
produced energy for the year 2010. The mean value of PI is very close to 85%, which indicates that, on
average, the PV systems are producing an annual energy that is 15% inferior to the reference system. The
PI observed in 2010 tends to be slightly higher for newer installations. The mean PI measured in 2010 for
PV systems installed in 2007, 2008 and 2009 yielded respectively 82.9%, 83.5% and 85.5%. Two main
causes probably explain that trend. First, the power of PV modules is known to decrease with time due to
the light soaking. Second, quality controls have been given a growing importance during these last years. It
was not possible to track PI values from previous works to compare them with the ones obtained in the
present study. To make possible a direct comparison using the more widely spread concept of PR, Fig. 8
shows its corresponding histogram. The mean value of PR is 76%. As a comparison, values of PR between
48% and 93% have been reported in other works[23,24].
The distribution of PI is nearly normal between values from 70% to 100%. It is left skewed, which physically
arises from the existence of PV systems suffering from major issues and thus showing PI values
abnormally low, while even a very good PV system can hardly have a PI much higher than 100%. The
skewness can be approximated through a Weibull distribution (at a confidence level of 95%, Anderson-
Darling goodness of fit = 1.452). The distribution of PR is more symmetrical, mainly because the influence
of cell’s temperature on the PR introduces wider variation among the PV systems installed in different
Fig. 8. Histogram of the performance ratio (above) and performance index (below) of the PV systems
analyzed. The distribution is nearly normal between a PI value of 70% and 100%. The distribution is left
skewed. The skewness is modeled through a Weibull distribution. The distribution of PR is more
symmetrical, mainly because the influence of cell’s temperature on the PR introduces wider variation
among the PV systems installed in different climatic conditions.
In order to look for the causes that explain the PI differences among the different PV residential systems,
an ANOVA was applied to the whole database. It did not allow associating significant variations of PI to the
nominal power of the installations, the inverter manufacturers or the installers. This failure to identify
significant trends does not imply the absence of differences. It simply means that the PI differences cannot
be statistically attributed to any of these parameters with a sufficient confidence level.
The ANOVA did however allow to establish strong evidence that the PV modules explain the majority of the
dispersion of PI (F=23.21 and P-Value <0.001). The results of this ANOVA for PV modules that are present
on at least 25 installations are detailed on table 3. Manufacturers’ names have been hidden under symbols
for confidentiality reasons. xSi stands for crystalline silicon; bcSi stands for back-contact silicon; HIT stands
for heterojunction with intrinsic thin layer; CIS stands for CuInSe2 based solar cell (thin film). Among the
results, it is possible to draw important observations about two PV modules technologies. On the one hand,
the PV systems equipped with the module tagged as “HIT1” show PI values higher than average. This
module is also the most represented on the PV systems of the database. On the other hand, the systems
equipped with the PV module tagged as “CIS” clearly show a PI pretty low respect to all the other groups.
Table 3: ANOVA on PV Modules present at least on 25 PV installations. N indicates the number of
installations. The ANOVA analysis on PV modules shows significant differences between the mean power
of several groups of PV modules.
Mean of PI
StDev of PI
Fig. 9 shows a boxplot that allows visualizing the PI variations among and within the groups of PV modules.
Fig. 9. Boxplot of performance index for PV modules present at least on 25 PV installations. The boxes
show the first, second and third quartiles, represented respectively by the lower, medium and upper
horizontal lines. The second quartile is also the median.
In order to estimate the real power of the PV modules, we assume that the losses due to the Balance of
System (BOS) are 10% higher than in the reference system. This assumption is supported by previous
works that describe the losses typically present at a PV system. The soiling losses typically account for
3%[19,20]. The average inverter has a yield 2% lower than the high quality inverter that equips the reference
system. PV generator mismatch and wiring losses can typically be 2% higher than in the reference
system. Shading can lead to important energy losses in some cases. The evaluation of shading losses is
particular to each project and often implies complex models. The shading losses were not simulated for
each PV system, but were instead estimated to 2% on average, which seems a reasonable hypothesis for
the typical residential PV systems in France. Other losses, such as the ones due to the availability of the
system, can account for 1%. Those losses can thus be estimated conservatively to account for 10% of
annual energy losses. As the mean value of PI is 85%, there is a 5% left that is probably due to a power
default in the PV modules.
Under those assumptions, it is possible to group the PV modules by manufacturer and to estimate the
deviation of their real power respect to their corresponding nominal power announced by the manufacturer.
Fig. 10 shows the result of this exercise for 51 different manufacturers of PV modules. It is worth
mentioning that the PV modules analyzed here have a mean exposure time of 2 years.
Fig. 10. Histogram of the deviation of the real power of the PV modules respect to their nominal power. On
average, the PV modules real power falls 4.9% below their corresponding nominal power.
The majority of the PV modules have a real power between 2% and 8% lower than their nominal power.
The presence of PV modules showing a real power higher than their nominal power corresponds to PV
modules delivered with positive power tolerances, or to a BOS better than the one considered in this
analysis, or a combination of both factors. Two kinds of averages can be used to characterize the
distribution as a whole. The first possibility is to give the same weight to the power deviations of each PV
module manufacturer (unweighted average). The second possibility, more representative of the state of the
art, is to weight the power deviations of each PV module manufacturer by the total power of its modules
present in the sample (weighted average). The unweighted average yields 6.3%, and the weighted average
yields 4.9%. The PV module that yields the best results is also the one that sold the best in 2010. It is a
module based on HIT technology. Some models of PV modules show poor quality. A relevant observation
concerns a PV module based on CIS technology, showing a mean real power up to 16% below the nominal
power. To investigate the reasons for such a low power, the PV systems equipped with this module have
been grouped by year of installation. On average, a loss of power of 5% per year has been observed on
these modules from 2007 to 2010. The low power is thus very probably due partly to an initial low power,
and partly to a light soaking degradation higher than for the other technologies. The multidimensional
ANOVA allowed verifying that those conclusions about the real power of PV modules are not affected by
other parameters of the installations, such as the inverters or installers.
The comments received from the users of BDPV indicate that the geographical origin of the PV modules
and inverters is often taken as an indicator of quality. In particular, they tend to consider as high quality the
PV components manufactured in their country or in Europe, while they often turn down PV modules
manufactured in China on the sole basis of their origin. Nevertheless, the analysis of the data of BDPV has
demonstrated that no clear correlation exist between the performance of PV modules and the country
where they were made. For example, several Chinese manufacturers present in the database perform
better than average, while several French PV modules manufacturers perform below average.
Other authors have reported the real nominal power of PV modules to be on average 5% inferior to the
nominal power stated by their manufacturer[26,29]. Globally, those differences between real power and
nominal power suggest that it is profitable to implement quality control procedures to verify and improve the
quality of PV systems[30,31].
4.4. Validation of the results against on-site measurements
The results of the present work have been compared to on-site measurements realized by the IES-UPM
during the last years on more than 200 MW of PV modules equipping solar plants in Spain, Italy and
France. Among the modules manufacturers present on more than 25 installations of the database of
BDPV, it was possible to identify 9 that were also measured by the IES-UPM. The relation between their
real and nominal powers has been estimated by both methods, and is shown on table 4. The difference is
generally lower than 2.3%. This is well within the uncertainties of those kinds of measurements. Additionally
to the uncertainties on the measurements and on the estimations used in both methods, other sources of
uncertainties are present from the fact that the PV modules that were rated here were not all installed at the
same time, nor under the same climatic conditions, which possibly implies that some modules have already
lost more power than others due to the light soaking process ongoing since they were exposed to outdoor
conditions. Such agreements can thus be considered as very satisfactory. It is important to note that these
conclusions are drawn only for PV modules present on at least 25 installations.
Table 4: Comparisons between real and nominal powers estimated in this work, and measured by IES-
UPM on solar plants in Spain, Italy and France.
Real vs Nominal
Real vs Nominal
5. Conclusions and discussions
The objective of this paper is to review the state of the art of residential PV systems in France, which is
done analyzing the operational data of 6868 PV systems. Although the available data relate to installations
in Metropolitan France, the results are of general interest to understand the state of the art of residential PV
The PV market in France developed towards residential PV systems as a consequence of limiting the most
interesting public financial support to systems of a maximum of 3 kWp. The PV industry (manufacturers of
PV modules and inverters) is dominated by a reduced number of actors, while an important fraction of
installations are realized by small installers, working at a regional scope.
On average, the PV systems produced in 2010 a net annual energy of 1163 kWh/kWp. As a whole, the
orientation of residential PV causes energy productions to be some 7% inferior to optimally oriented PV
systems. These losses due to orientation are generally low enough to ensure that the PV systems installed
on buildings are a viable alternative to solar plants optimally oriented.
The quality of the PV systems is quantified using the performance ratio (PR), and the performance index
(PI). After a mean exposure time of 2 years, the mean value of PR is 76% and the mean PI is 85%, which
implies that the typical real PV system produces 15% less than a reference PV system. On average, the
real power of the PV modules falls 4.9% below their corresponding nominal power announced on the
manufacturer’s datasheet. A brief analysis by PV modules technology has led to relevant observations
about two technologies. On the one hand, the PV systems equipped with HIT modules show performances
higher than average. On the other hand, the systems equipped with the CIS modules show a real power
that is 16 % lower than nominal value.
This work would never have been possible without the thousands of altruistic people who generously and
freely provided us the data corresponding to the energy production of their PV system. Catherine Praile,
owing to her programmer talents, was absolutely essential to the success of this work. Eduardo Lorenzo
brought invaluable feedback and insight.
 Statistiques EDF, Installations de production raccordées au réseau géré par ERDF à fin mars 2011.
 BDPV: Base de Données PhotoVoltaïque (http://www.bdpv.fr/).
 Jahn, U. and Nasse, W., Operational performance of grid-connected PV systems on buildings in Germany.
Progress in Photovoltaics: Research and Applications, 12: 441–448. doi: 10.1002/pip.550, 2004.
 Oozeki, T., Yamada, T., Otani, K., Takashima, T. and Kato, K., Performance trends in grid-connected photovoltaic
systems for public and industrial use in Japan. Progress in Photovoltaics: Research and Applications, 18: 596–602.
doi: 10.1002/pip.992, 2010.
 Wiles J.C., Brooks B. and Schultze B., PV installations, a progress report, IEEE Photovoltaic Specialists
 Dincer F., The analysis on photovoltaic electricity generation status, potential and policies of the leading countries
in solar energy, Renewable and Sustainable Energy Reviews, Volume 15, Issue 1, January 2011, Pages 713-720,
ISSN 1364-0321, DOI: 10.1016/j.rser.2010.09.026.
 IEA, PVPS Task 13: Performance and Reliability Of Photovoltaic Systems, in progress (http://www.iea-
 SoDa (http://www.soda-is.com/).
 PVGIS (http://re.jrc.ec.europa.eu/pvgis/).
 Erbs D.G., Klein S.A. and Duffie J.A., Estimation of the diffuse radiation fraction for hourly, daily and monthly-
average global radiation, Solar Energy, Elsevier, 1982.
 Liu B. and Jordan R.C., The interrelationship and characteristic distribution of direct, diffuse, and total solar
radiation, Solar Energy 1960;4:1–9.
 Hay J.E., Calculating solar radiation for inclined surfaces: Practical approaches, Solar Energy, Elsevier, 2003.
 Narvarte L. and Lorenzo E., Tracking and ground cover ratio. Progress in Photovoltaics: Research and
Applications, 16: 703–714. doi: 10.1002/pip.847, 2008
 Marion B., A method for modeling the current–voltage curve of a PV module for outdoor conditions, Progress in
Photovoltaics: Research and Applications, Volume 10, Issue 3, Pages 205–214, 2002.
 Skoplaki E., Palyvos J.A., On the temperature dependence of photovoltaic module electrical performance: A
review of efficiency/power correlations, Solar Energy, Elsevier, Volume 83, Issue 5, 2009, Pages 614-624.
 Simon M., Meyer E.L., The effects of spectral evaluation of c-Si modules, Progress in Photovoltaics, 2011.
 Betts T, Jardine C, Gottschalg R, Infield D, Lane K. Impact of spectral effects on the electrical parameters of
multijunction amorphous silicon cells. IN: Proceedings of 3rd World Conference on Photovoltaic Energy Conversion,
Osaka, Japan, 11-18 May, 2003, 2, pp. 1756-1759".
 Schmid J and Schmidt H, Inverters for photovoltaic systems, Proceedings 5th Contractor’s Meeting of the
Photovoltaic Demonstration Projects, Ispra, 1991.
 Hammond R, et al" by "Hammond, R. Srinivasan, D. Harris, A. Whitfield, K. Wohlgemuth, J., Effects of soiling
on PV module and radiometer performance, AZ Photovoltaic Specialists Conference, 1997.
 Martin N., Ruiz J.M., Annual angular reflection losses in PV modules, Progress in Photovoltaics, 2006.
 Behnke, Erdman, and Whitaker Engineering, Inc., Commonwealth Energy Biogas/PV Mini-Grid, Renewable
Resources Program, Interim Exposure and Operations Report for the Comprehensive Large PV System Comparison,
 Martínez-Moreno F., Lorenzo E., Muñoz J., Moretón R., Testing of PV arrays in large grid-connected PV plants
installed in Spain, 23rd EUPVSEC, 2008.
 Decker B, Jahn U. Performance of 170 grid connected PV plants in northern Germany—analysis of yields and
optimization potentials. Solar Energy 1997;59:127–33.
 Munro D.K., Blaesser G., The performance of PV systems and components in the THERMIE programme,
Renewable Energy, Volume 5, Issues 1-4, Climate change Energy and the environment, August 1994, Pages 172-
178, ISSN 0960-1481, DOI: 10.1016/0960-1481(94)90368-9.
 Muñoz, J., Martínez-Moreno, F. and Lorenzo, E., On-site characterisation and energy efficiency of grid-connected
PV inverters. Progress in Photovoltaics: Research and Applications, 19: 192–201. doi: 10.1002/pip.997, 2011
 Martínez-Moreno, F., Lorenzo, E., Muñoz, J. and Moretón, R., On the testing of large PV arrays. Progress in
Photovoltaics: Research and Applications, 19: n/a. doi: 10.1002/pip.1102, 2011.
 Martinez-Moreno F., Munoz J., Lorenzo E., Experimental model to estimate shading losses on PV arrays, Solar
Energy Materials and Solar Cells, Volume 94, Issue 12, December 2010, Pages 2298-2303, ISSN 0927-0248, DOI:
 Bletterie B., Sensitivity of photovoltaic inverters to voltage sags — Test results for a set of commercial products, e
& i Elektrotechnik und Informationstechnik, Volume 122, Number 10, 349, 2007.
 Fraunhofer-Institut Für Solare Energysysteme ISE, Annual Report, Achievements and Results, 2007.
 Caamaño E., Lorenzo E., Zilles R., Quality control of wide collections of PV modules: lessons learned from the
IES experience, Progress in Photovoltaics: research and applications 7, 137-149, 1999.
 Coello J, Galindo JL, Carames M, Carreño R, Quality control during the supply of PV modules: Fundamental key
to guarantee the profitability of PV installations, Proceedings 23rd European Photovoltaic Solar Energy Conference,