Conference PaperPDF Available

Performance of Distributed PV in the UK: A Statistical Analysis of Over 7000 Systems

Authors:

Abstract and Figures

In June 2015, the UK fleet of solar photovoltaic (PV) systems reached 7.8 GWp of capacity, but there are wide gaps in our understanding of the performance of these systems, which has lead to the conservative limit of 10 GWp being imposed on UK PV capacity by the Department of Energy and Climate Change. Here we present the results of a statistical analysis of real world UK PV systems which donate data to the Microgen Database, of which there are over 7000. The mean yearly-integrated Performance Ratio (PR) of domestic scale UK PV is 83% with a standard deviation of 7%. By considering yearly-integrated PR, we have shown that 4.1 % of systems suffered long-term underperformance relative to their nominal efficiencies during 2013. The mean degradation rate for crystalline Silicon-based PV systems in the UK is-0.8 ± 0.1% per year. The state-of-the-art of UK PV, in terms of technology, manufacturing, and installation-standards, is found to have increased by 1% per year between 2002 and 2013.
Content may be subject to copyright.
31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
PERFORMANCE OF DISTRIBUTED PV IN THE UK: A STATISTICAL ANALYSIS OF OVER 7000
SYSTEMS
Jamie Taylor1,*, Jonathan Leloux2, Lisa M.H. Hall1, Aldous M. Everard1, Julian Briggs1, Alastair Buckley1
1 Sheffield Solar, University of Sheffield, Hicks Building, Hounsfield Road, Sheffield S3 7RH, UK
2 Instituto de Energía Solar, Universidad Politécnica de Madrid, Spain
* Corresponding Author jamie.taylor@sheffield.ac.uk
ABSTRACT: In June 2015, the UK fleet of solar photovoltaic (PV) systems reached 7.8 GWp of capacity, but there
are wide gaps in our understanding of the performance of these systems, which has lead to the conservative limit of 10
GWp being imposed on UK PV capacity by the Department of Energy and Climate Change. Here we present the results
of a statistical analysis of real world UK PV systems which donate data to the Microgen Database, of which there are
over 7000. The mean yearly-integrated Performance Ratio (PR) of domestic scale UK PV is 83% with a standard
deviation of 7%. By considering yearly-integrated PR, we have shown that 4.1 % of systems suffered long-term
underperformance relative to their nominal efficiencies during 2013. The mean degradation rate for crystalline Silicon-
based PV systems in the UK is -0.8 ± 0.1% per year. The state-of-the-art of UK PV, in terms of technology,
manufacturing, and installation-standards, is found to have increased by 1% per year between 2002 and 2013.
Keywords: System Performance, Degradation, Small Grid-connected PV Systems
1 INTRODUCTION
In June 2015, the UK fleet of solar photovoltaic (PV)
systems reached 7.8 GWp of capacity [1]. Previous works
have studied the performance of some PV systems in the
UK [2], but there are few publications [3] which study an
ensemble of systems representative of the UK fleet,
leading to gaps in our knowledge of real-world
performance of distributed PV in the UK. This work
explores the real-world generation of over 7000 distributed
PV systems from the Microgen Database (MgDB) [4], in
order to characterise and quantify performance. The
resulting statistics will help to inform both academia and
industry, whilst also informing Government policy with
respect to PV. We explore areas of key interest to
stakeholders such as performance ratio (PR), state-of-the-
art and degradation.
These results are highly relevant to the decision
making process undertaken by policy makers in the UK
due to the implications for cost analysis of incentives and
ensuring effective integration into the electricity network.
This is especially true at the time of writing, since the
Department of Energy and Climate Change (DECC) for
the UK Government recently opened a consultation on a
review of Feed-in Tariffs (FITS) for micro-generation PV
to take place in January 2016 [5].
Several thousand PV systems have been monitored
since 2010, with some systems’ historic data spanning
over 7 years. The performance of UK PV has been
characterised by deriving statistics regarding key
performance metrics; PR at a monthly and yearly
integration and PR at standard test conditions (PR@STC).
These metrics then provide the basis for assessing
improvements in the state-of-the-art of PV as well as
deriving an indicative measured degradation rate for PV
performance in the UK.
2 DATA AND METHOD
2.1 Data
Distributed PV generation data is collected via the
MgDB website [4], 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.
Figure 1; Map of the MgDB systems used in this
analysis.
The complete dataset of MgDB comprises PV
generation data from more than 7000 PV systems across
UK (Figure 1), at various temporal resolutions (typically
10-min, 30-min, daily or monthly), with historic data
spanning up to seven years [4], although most of the PV
systems were installed after 2011. The dataset is supplied
by a combination of homeowners and commercial sources
and includes both domestic and commercial scale
installations between 0.7 and 69 kWp with a wide range of
orientation and tilt angles. The data from the MgDB has
been subjected to rigorous checks and validations in order
to isolate and remove as much erroneous data as possible.
The standard set of filters employed prior to analyses is:
Use only single array systems since generation
31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
data cannot be decomposed into constituent
arrays.
Use only systems within the UK. This is
necessary since the MgDB website accepts
systems from anywhere in the world, although
in reality only a very small proportion lies
outside of UK.
Use only systems with 
 and 
.
In some cases system data is investigated manually to
verify to a good degree of confidence that the data should
be removed, for example when considering systems whose
orientation or tilt appears incorrect. After the reading
requirements and system validation has been carried out
4369 systems remain, which are analysed in this study.
Data from the MgDB is prone to human errors on the part
of the donor, for example, entering incorrect system
parameters such as orientation, tilt or installed capacity.
Some of these errors lead to outliers in the distribution of
PR and/or PR@STC which can skew non-robust statistics
such as the mean, , and standard deviation, . It is
therefore crucial that we are able to identify and isolate
them from the analysis. A simple and reliable method for
removing outliers from a symmetrical distribution is
Tukey’s method [6], which uses the outlier limits in
Equation (1).




Tukey’s method proves useful when the aim is to
remove all outliers which do not form part of the
symmetrical distribution, which in this context
corresponds to systems that are performing correctly i.e.
no underperformance. This is desirable when we
investigate the degradation since we want our result to be
representative of a fully functioning PV system i.e. we are
only interested in degradation and in the case of
underperforming systems we cannot distinguish between
the degradation and underperformance due to other
factors.
When analysing the distribution of yearly-integrated
PR, it is desirable to include the results for under-
performing systems whilst excluding any outliers due to
erroneous data and complete failures. This is complicated
by the fact that the distribution takes the shape of a Weibull
distribution [7] [8] which is non-symmetric and features a
long tail at lower values. To achieve this, we employ a
method developed specifically in the context of PV fault
detection [9] whereby the upper limit is the median, ,
plus  and the lower limit is  minus 
(Equation (2)). The statistic  is the standard deviation
of all values above the median, that is, the standard
deviation of the observed normal part of the distribution.


2.2 Irradiation
Monthly Global Horizontal Irradiation (GHI) has been
interpolated at each of the sites from the UK Met Office
(UKMO) ground based pyranometers [10] using an
inverse distance weighted interpolation as per the
methodology documented by Colantuono et al. [3]. As
with Colantuono et al., the exponent of the inverse distance
is chosen to minimise the mean error across all UKMO
stations using leave-one-out cross-validation (LOOCV).
By applying LOOCV to 96 months (2011-2014) of
interpolated monthly irradiation data, the mean absolute
percentage error (MAPE) of this interpolation method in
UK has been calculated as 5.0%, whilst the mean
percentage error (MPE) of all months, 0.1%, reveals
negligible bias overall. The resulting overall root mean
square error (RMSE) is 4.5%. For this LOOCV we use the
5%-trimmed-mean in place of the mean to account for and
remove the effect of highly localised weather conditions,
which are circled in red in Figure 2. These uncertainty
estimates are in line with those reported by Colantuono et
al. We have calculated the 5% trimmed MPE for each
season during the four year period and find the range to be
between 0.1 and 0.2%, indicating that this interpolation
method is in general resilient to bias in all seasons. The
MAPE increases in winter relative to summer, with values
of 4, 4, 5 and 7% for spring, summer, autumn and winter
respectively.
Figure 2; Distribution of errors for interpolated
irradiation using LOOCV across 96 months.
The interpolated monthly GHI is decomposed into
direct and diffuse components according to Page [11]. The
direct and diffuse components are then transposed to the
inclined plane and summed to give the Global Tilted
Irradiation (GTI) using Klein & Theilacker [12].
2.3 Monthly and yearly-integrated PR
Performance Ratio (PR) is widely used metric for
comparing relative performance of PV systems whose
design, technology and location differ [13]. PR is defined
in Equation (3):

 
 
Where  and  are the achieved efficiency
and nominal efficiency (according to the manufacturer) of
the system respectively (dimensionless); is the energy
generated by the system (kWh) and  is the irradiation
incident in the plane of the array (kWh). The achieved
efficiency must be calculated over some arbitrary period.
In the case of a yearly period, we refer to the PR as the
yearly-integrated PR in order to distinguish it from the
mean of the monthly PR across all months in the year,
which is not studied here.
31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
We have analysed the distribution of the yearly-
integrated PR on a histogram after applying the outlier
removal technique described by Equation (1) and have
fitted several continuous distributions in order to quantify
the shape of the distribution and offer reproducibility. We
have graphed mean monthly PR across all systems in order
to demonstrate seasonal variability.
2.4 Monthly PR@STC
PR fails to take into account the module efficiency
response to variations in module temperature and
irradiance intensity. In the UK these factors are highly
seasonal and as a result there is a significant seasonal and
inter-annual variation in the measured PR. The so-called
Performance Ratio at Standard Test Conditions
(PR@STC) attempts to correct for these effects by
introducing correction terms to the PR calculation.
According to the US National Renewable Energy
Laboratory (NREL) [14], PR@STC is given by Equation
(4):





 
Where  is the uncorrected monthly PR; is the
temperature coefficient of power for the installed modules
(%/°C, negative in sign); 
is the cell temperature at
STC (25 °C);
 is the irradiance-weighted mean cell
temperature for the month (°C, see Equation (5)); is a
parameter describing the reduction in efficiency due to
decreased irradiance (0.031 for crystalline Silicon cells
based on 
 );  is the mean irradiance-
weighted irradiance in the plane of the array for the month
(W/m2) and is the irradiance under STC (1000 W/m2).
The irradiance-weighted mean cell temperature,
, is
given by Equation (5):
 


Where  and  are the cell temperature
and irradiance respectively at hour ; and is the
summation over all hours in the month.
The cell temperature at each hour is estimated using the
Sandia PV Array Performance model [15], which takes
irradiance, ambient temperature, wind speed and module
parameters as inputs. Conveniently, the Sandia model is
available in the PV_LIB Matlab library, made available
as an open-source project by the Sandia National
Laboratories PV Performance Modelling Collaborative
[16]. The module parameters determine the heat transfer
coefficients, , and , according to the type of array;
for these analyses we use the “Glass/cell/glass” and
“Close-roof mount” values of -2.98, -0.0471 and 1°C
respectively, as recommended by King and Boyson [15].
Since
 requires hourly irradiance, ambient
temperature and wind speed data, we have only considered
systems close to (within 20 km of) UKMO weather
stations, where this data is readily available.
2.5 System level performance degradation
Monthly PR@STC during the months April to
September has been shown to be stable (Figure 6 and 7)
and so it provides a useful benchmark for year-on-year
comparisons of performance. By analysing the monthly
PR@STC during these months on a per system basis over
several years, we have derived an indicative value for the
relative system level performance degradation of UK
distributed PV during the first years of operation. This is
achieved by first normalising the April-September
PR@STC of each system to the earliest value and then
fitting a straight line to the data with a fixed intercept of 1.
Here we present a histogram of the resulting degradation
rates and derive an average rate using robust statistics. In
order to be included, a system must have at least 5 monthly
PR@STC data points spanning at least 3 years.
2.6 State of the art
We have assessed the improvement in the state of the
art of the PV systems installed in the UK by analysing
yearly integrated PR as a function of installation date. A
linear fit on the resulting graph is used to extract the rate
of improvement in the state of the art of UK distributed
PV. Because the date of installation was not available for
all of the PV systems of the MgDB, we have used the
production start date of the installed modules as a proxy
for installation date. In doing so, we have accurately
represented improvements in state of the art due to the
supply-chain and manufacturing process, but will have
introduced some uncertainty with regards to installation
standards since there may be some lag in the correlation
between production start date and install date.
3 RESULTS AND DISCUSSION
3.1 Monthly and yearly-integrated PR
Figure 3; Histogram of yearly-integrated PR across all
years of data.
The distribution of yearly-integrated PR is presented
in Figure 3. The shape is characterised by a Weibull type
distribution whereby the normal uncertainty in the PR
calculation is superimposed with a long tail to lower
values, corresponding to underperforming systems. We
also fit a Johnson Su distribution in order to provide
reproducibility. The mean yearly-integrated PR is 83.33%
with a standard deviation of 6.68% and a standard error of
0.08%. The median yearly-integrated PR is 84.60%. The
boxplot in Figure 3 displays the Tukey outlier limits
discussed earlier and demonstrates why such limits are
effective in removing underperforming systems, i.e.
systems in the tail of the distribution. The mean is higher
than the values reported recently across Europe [8] [17]
31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
[18]. Some of this discrepancy may be due to uncertainties
in the interpolation of GHI at the location of the system
since each source of irradiation data will be subject to
different errors. For example, irradiation data from the
Climate Monitoring Satellite Application Facility (CM-
SAF) has been shown to be subject to bias errors of more
than +15% [19], with a relative mean bias error of +6.2%
in the UK. Another potential source of discrepancy is the
transposition of GHI to GTI.
Figure 4; Histogram of yearly-integrated PR in 2013.
Data highlighted in red correspond to systems experience
long term underperformance according to our 
 rule.
By considering systems with yearly-integrated PR less
than , we can determine the proportion of
systems that are deemed to be underperforming with
respect to their peers. Figure 4 presents the distribution of
yearly-integrated PR for all systems in 2013 once
erroneous data has been removed according to the
 rule discussed earlier. The normal part of the
distribution has been fitted using  and ,
with underperforming systems highlighted in red. Of the
4181 systems, 171 are deemed to have underperformed in
2013, equivalent to 4.1%.
Figure 5; Histograms of monthly PR by year and season
for 2012-2014.
In figure 5 we see the distribution of monthly PR
broken down into seasons across 3 years of data. It is clear
from the spread in the histograms and the standard
deviations that the PR is less variable in the summer
months than in winter, with spring and autumn falling
somewhere in between. Figure 5 also reveals the year-on-
year variation to be significantly less during summer than
in winter, with ranges of 1.8% and 7.5% respectively. This
trend is consistent with that reported in other European
countries with similar climate [8] [17] [20]. For these
reasons, monthly PR is of limited use as a means to
monitor PV systems for underperformance and faults.
3.2 Monthly PR@STC
As with PR, the PR@STC is more variable in winter
than in summer, making it of limited use as a monitoring
tool (Figure 6). With PR@STC the dip in summer due to
increased temperatures is less pronounced [7] whilst a
pronounced dip in winter indicates underperformance
during these months. The main driving factor for this
underperformance is thought to be increased shading as a
result of lower solar elevation angles, but the drop in
efficiency of inverters at lower generation levels may also
contribute.
Figure 6; Mean monthly PR@STC for 2012-2014.
Figure 7 compares the standard deviation in the
measured PR@STC for each month over all available
systems and all available years. In order to eliminate the
effect of underperforming systems, PR and PR@STC
values were removed according to the Tukey outlier limits.
In general the PR@STC shows less variance and is
therefore an improvement over PR, but the increased
standard deviation during winter months relative to
summer means it is still problematic. Figure 7
demonstrates that PR@STC is consistently accurate from
April to September inclusive and so we choose this period
to assess year-on-year degradation.
Figure 7; Standard deviation of the PR and PR@STC for
each month across all years and all systems.
3.3 System level performance degradation
Before analysing the distribution of degradation rates,
we remove those that fall outside Tukey outlier limits. These
systems are badly fit by straight line, most probably because
of temporary faults or down time. We weight the
distribution according to the number of years of data
available for each system, since this will effectively weight
towards more reliable fits. The resulting degradation rate
will only be representative of the MgDB sample, which is
31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
overwhelmingly comprised of crystalline Silicon
technology.
The skew (Pearson's moment coefficient of skewness)
of -2.5 indicates definite shift in distribution towards
negative values. The mean degradation is -0.8% per year
with a standard error of 0.1%, whilst the median is -0.5%
per year. These values are in good agreement with literature
[21] [22]. The 95% confidence interval for the mean lies
between -0.6 and -1.0 % per year.
Figure 8; Histogram of degradation rates calculated for
all systems from PR@STC. It is important to note that the
spread in the distribution should be attributed to the
uncertainty in the PR leading to uncertainty in the
degradation.
It is worth noting that the resulting mean differs slightly
depending on whether PR@STC outliers (underperforming
systems) are included or excluded. The mean without
outliers is -0.8%, whilst the mean with outliers is -0.9% per
year. It is not possible to discern whether the
underperforming systems are such because of increased
degradation, or whether the nature of the fault that caused
the underperformance has then led to increased degradation.
It is therefore appropriate to exclude the outliers, and the
resulting mean can be seen as representative of the
degradation experienced by systems that have not developed
any specific faults.
3.4 State of the art
Figure 9; Boxplots of yearly-integrated PR in 2014
against production start year of the installed modules.
Orange line shows a linear best fit with gradient of 1%
per year. The orange shaded area shows the 95%
confidence interval for the linear fit.
Figure 9 shows the 2014 yearly-integrated PR of 866
systems against the production start year of the installed
modules. The gradient of the linear fit implies that the
system performance has increased by 1.1% per year. A
similar fit to 2012 and 2013 yearly-integrated PR data
yields gradients of 0.97 and 0.95% from 348 and 890
systems respectively. We use the mean of these values
weighted according to the number of systems each year in
order to calculate the improvement in the state-of-the-art
of UK PV to be 1.0% per year. This is lower than
previously reported [7], but is more robust thanks to a
thorough outlier removal process and the inclusion of
more than one year’s yearly-integrated PR data. The 95%
confidence interval for the gradient is 0.68 to 1.4% per
year. The increase in performance is thought to be driven
by improvements in the manufacturing and distribution
process as well as improvements in installation standards.
The implication for the UK PV industry is that the state-
of-the-art of small-scale PV has improved by 10% over the
last decade.
4 CONCLUSION
We have presented a detailed statistical analysis of a
large ensemble of UK domestic scale PV, focusing on
statistics of crucial importance to policy makers, industry
and academia. The yearly-integrated PR in the UK is
found to be higher than the equivalent statistic reported in
other European countries [8] [17] [18]. The discrepancy
with European studies requires further investigation as it
is not clear whether this arises from genuine physical
phenomenon or some undiagnosed source of bias in this or
the other studies considered.
We have identified that 4.1% of PV systems suffered
long term underperformance during 2013, which is also
indicative of the underperformance rate in other years.
We have presented a mean degradation rate for
crystalline Silicon of -0.8 ± 0.1% per year in the UK,
which is in good agreement with values reported
elsewhere [21] [22].
We have established an improvement in the state of the
art of the UK PV industry of 1% per year between 2002
and 2013, amounting to a 10% improvement in the last
decade.
5 REFERENCES
[1]
UK Office of Gas and Electricity Markets
(OFGEM), "Feed-in Tariff Installation Report 30
June 2015," 06 07 2015. [Online]. Available:
https://www.ofgem.gov.uk/environmental-
programmes/feed-tariff-fit-scheme/feed-tariff-
reports/installation-reports. [Accessed 11 09 2015].
[2]
E. A. Essah, A. R. Arguelles and N. Glover,
"Assessing the performance of a building
integrated BP c-Si PV system," Renewable Energy,
vol. 73, pp. 36-45, 2014.
[3]
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.
[4]
Sheffield Solar, "Microgen Database," Sheffield
Solar - University of Sheffield, [Online]. Available:
http://www.microgen-database.org.uk/.
[5]
UK Department of Energy and Climate Change
(DECC), "Consultation on a review of the Feed-in
Tariffs scheme," 27 08 2015. [Online]. Available:
https://econsultation.decc.gov.uk/office-for-
31st European Photovoltaic Solar Energy Conference and Exhibition, September 2015, Hamburg
renewable-energy-deployment-ored/fit-review-
2015. [Accessed 15 09 2015].
[6]
J. W. Tukey, Exploratory Data Analysis, Addison-
Wesley, 1977.
[7]
J. Taylor, J. Leloux, A. Everard, J. R. C. Briggs and
A. Buckley, "Monitoring thousands of distributed
PV systems in the UK: Energy production and
performance," in PVSAT-11, Leeds, 2015.
[8]
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.
[9]
J. Leloux, L. Navarte, A. Luna and A. Desportes,
"Automatic fault detection on BIPV systems
without solar irradiation data," in 29th European
Photovoltaic Solar Energy Conference and
Exhibition, Amsterdam, 2014.
[10]
Met Office, NCAS British Atmospheric Data
Centre, "Met Office Integrated Data Archive
System (MIDAS) Land and Marine Surface
Stations Data (1853-current)," 2012. [Online].
Available:
http://catalogue.ceda.ac.uk/uuid/220a65615218d5c
9cc9e4785a3234bd0. [Accessed 11 09 2015].
[11]
J. Page, "The estimation of monthly mean values of
daily total short-wave radiation on vertical and
inclined surfaces from sunshine records for
latitudes 40ºN40ºS," in Proc. UN Conf. on New
Sources of Energy, 1961.
[12]
J. A. Klein and J. C. Theilacker, "An Algorithm for
Calculating Monthly-Average Radiation on
Inclined Surfaces," Journal of Solar Energy
Engineering, vol. 103, pp. 29-33, 1981.
[13]
US National Renewable Energy Lab (NREL),
"Performance parameters for grid-connected PV
systems," in Photovoltaic Specialists Conference,
2005. Conference Record of the Thirty-first IEEE,
Florida, 2005.
[14]
S. Kurtz, J. L. B. Cruz, E. Riley and C. Hansen,
"Weather-Corrected Performance Ratio," April
2013. [Online]. Available:
http://www.nrel.gov/docs/fy13osti/57991.pdf.
[Accessed 13 January 2015].
[15]
D. L. King, W. E. Boyson and J. A. Kratochvill,
"Photovoltaic Array Performance Model," Sandia
National Laboratories, 2004.
[16]
Sandia National Laboratories, "PV Performance
Modelling Collaborative (PVPMC)," 2014.
[Online]. Available: https://pvpmc.sandia.gov/.
[Accessed 17 09 2015].
[17]
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.
[18]
A. Woyte, "Monitoring of photovoltaic systems:
good practices and systematic analyses," in 29th
European Photovoltaic Solar Energy Conference
and Exhibition, Amsterdam, 2014.
[19]
T. Huld, R. Muller and A. Gambardella, "A new
solar radiation database for estimating PV
performance in Europe and Africa," Solar Energy,
vol. 86, p. 18031815, 2012.
[20]
J. Leloux, R. Moreton, L. Navarte, J. Taylor, D.
Trebosc and A. Desportes, "Monitoring 30,000 PV
Systems in Europe: Performance, Faults and
Stateof the Art," in 31st European Solar Energy
Conference and Exhibition (EU PVSEC 2015),
Hamburg, 2015.
[21]
D. C. Jordan and S. R. Kurtz, "Photovoltaic
Degradation Rates An Analytical Review,"
Progress in Photovoltaics: Research and
Applications, vol. 21, no. 1, pp. 12-29, 2013.
[22]
B. Muller, L. Hardt, A. Armbruster, K. Kiefer and
C. Reise, "Yield Predictions for Photovoltaic
Power Plants: Empirical Validation, Recent
Advances and Remaining Uncertainties," in 29th
European PV Solar Energy Conference and
Exhibition (EU PVSEC 2014), Amsterdam, 2014.

Supplementary resource (1)

... Additionally, the weighted temperature for all PV sites in the UK is equal to 11.2 °C, while it is equal to 21.4 °C for the Australian PV sites. The weighted temperature is calculated using Equation (1) [26]: The PV sites have been categorized in two groups: the first group contains PV sites A, B and C (located in the UK), whereas the second group consist of PV sites D, E and F (located in Australia). The solar irradiance (G) and ambient temperature (T) play major role on the performance and annual energy production for the PV panels. ...
... Additionally, the weighted temperature for all PV sites in the UK is equal to 11.2 • C, while it is equal to 21.4 • C for the Australian PV sites. The weighted temperature is calculated using Equation (1) [26]: ...
Article
Full-text available
This article presents the analysis of degradation rate over 10 years (2008 to 2017) for six different photovoltaic (PV) sites located in the United Kingdom (mainly affected by cold weather conditions) and Australia (PV affected by hot weather conditions). The analysis of the degradation rate was carried out using the year-on-year (YOY) degradation technique. It was found that the degradation rate in the UK systems varies from −1.05% and −1.16%/year. Whereas a higher degradation ranging from −1.35% to −1.46%/year is observed for the PV systems installed in Australia. Additionally, it was found that in the Australian PV systems multiple faulty PV bypass diodes are present due to the rapid change in the ambient temperature and uneven solar irradiance levels influencing the PV modules. However, in cold weather conditions (such as in the Northern UK) none of the bypass diodes were damaged over the considered PV exposure period. Furthermore, the number of PV hot spots have also been observed, where it was found that in the UK-based PV systems the number of hot spotted PV modules are less than those found in the Australian systems. Finally, the analysis of the monthly performance ratio (PR) was calculated. It was found that the mean monthly PR is equal to 88.81% and 86.35% for PV systems installed in the UK and Australia, respectively.
... Based on the analysis done by J. Taylor et al [23], it was found that the average annual degradation rate of 7000 PV modules in the UK is equal to -0.8%/year. Therefore, we have used this rate to reintroduce the CDF plots for the next 10 years or 20 years of the PV exposure, hence the plots shown in Figs. ...
... Figs. 7(c-d) presents the output results of the CDF plot for the PPL analysis for all hot-spotting categories including PV modules at their 21 to 30 operation. The used degradation rate is equal to -0.8%/year as suggested by [23]. A brief comparison for all obtained PPL thresholds are shown in Table I. ...
Article
In this paper, a novel Photovoltaic (PV) hot-spotting fault detection algorithm is presented. The algorithm is implemented using the analysis of 2580 polycrystalline silicon PV modules distributed across the UK. The evaluation of the hot-spots is analysed based on the cumulative density function (CDF) modelling technique, whereas the percentage of power loss (PPL) and PV degradation rate are used to categories the hot-spots into eight different categories. Next, the implemented CDF models are used to predict possible PV hot-spots affecting the PV modules. The developed algorithm is evaluated using three different PV modules affected by three different hot-spots. Remarkably, the proposed CDF models precisely categorize the PV hot-spots with high-rate of accuracy almost above 80%.
... While there is increasing evidence about the economic, energy system and environmental benefits of installing rooftop PV systems, there is also research which finds that some systems underperform [8][9][10][11]. In the UK, it was found that 4.1% of systems suffered long-term underperformance relative to their nominal efficiencies [9]. Other studies [10,11] identified dozens of fault types a range of PV system components including modules, connection lines, converters, and inverters, each leading to serious underperformance in system efficiency and energy yield. ...
... While there is increasing evidence about the economic, energy system and environmental benefits of installing rooftop PV systems, there is also research which finds that some systems underperform [8][9][10][11]. In the UK, it was found that 4.1% of systems suffered long-term underperformance relative to their nominal efficiencies [9]. Other studies [10,11] identified dozens of fault types a range of PV system components including modules, connection lines, converters, and inverters, each leading to serious underperformance in system efficiency and energy yield. ...
Chapter
Solar PV systems have become common-place in many cities and regions, and is a core technology in purpose-built low-energy homes, but evidence is emerging that in many cases electricity output may be significantly lower than expected. Information from in-home energy monitoring systems, interviews and informal discussions with residents has shed some light on the experiences and issues faced by the end-user, particularly those associated with operating a solar PV system to achieve a low-carbon lifestyle. Case studies of residents in different ownership and income situations, and from three distinct housing developments in Australia and England are used to highlight end-user experiences. The study finds that the residents face a range of issues including the initial sizing and commissioning, a lack of solar knowledge and expected generation performance, as well as regulatory barriers that limit the opportunity to upgrade system size.
... Van Sark et al. have published an overview of PV performance assessment by practical examples (van Sark et al., 2017) where PV system metadata is available, while solar irradiation is obtained from meteorological stations in vicinity of the system. Similarly, the evaluation studies of large fleet of PV systems in France , Belgium , UK (Leloux, 2015;Taylor et al., 2015) and USA (Bolinger et al., 2020) were done on systems with the whole metadata information, while irradiation data was derived from national meteorological agencies or reanalysis databases. Martín-Martínez(Martín-Martínez et al., 2019) reported an evaluation of selected large PV systems in Spain where all required data was available. ...
Article
Full-text available
Operators of photovoltaic (PV) systems, especially the small ones, monitor only the produced energy output, since they are not equipped with a meteorological station, or there is none in the neighbourhood. When evaluating a large number of PV systems across a region or country, access to the data due to privacy restrictions is usually very limited. To resolve both aspects, we report on a methodology to evaluate the performance of photovoltaic systems where only produced energy data and rated power of the PV system are available and all other data, including orientation and inclination angles, and the exact location are missing. We developed a novel algorithm for automatic detection of orientation/inclination angles and used the recently developed Typical Daily Profiles (TDP) methodology for a fast and accurate evaluation of PV systems. TDPs are representative hourly profiles of the observed parameter (e.g., power, irradiance and temperature) on a monthly basis. For the PV system degradation rate extraction, year-to-year TDPs are used. The case study of 957 PV systems in Slovenia in the period 2015-2019 reveals an average PV system performance ratio exceeding 85% and an average PV system rated power degradation rate of − 0.7% per year.
... The data for Spain covered mostly large-scale PV plants versus rooftop systems for the other countries. An analysis of the performance ratio of 7000 solar-PV systems in the UK using data from 2002 and 2013 established a typical performance ratio of 83% as the mode of the data distribution across all systems [48]. An analysis for Germany evaluated the PR for 100 rooftop solar-PV systems and found a median PR of 84% [49]. ...
Article
Full-text available
The amount of energy striking the earth’s surface in one hour is higher than global annual societies energy use, yet the fraction of incoming solar radiation that can be harvested is significantly constrained. A global grid-cell methodology was adopted to assess the available global solar energy potential taking into account four constraints: land-use, solar irradiation, solar-to-electric technology, and net energy. Net energy is the amount of energy that is delivered to end-users, after subtraction of the energy inputs needed for capital infrastructure and operation. Both photovoltaic and concentrated solar power technologies are considered. The resulting constrained solar potential worldwide was estimated at 1098 exajoules per year, of which 98%, 75%, and only 15% can be extracted if the system needs to deliver an energy return on energy invested set at 5, 7.5, and 9, respectively. The resulting global solar potential is substantially lower than most previous estimates. Depending on how high the energy return needs to be relative to the energy investment needed to maintain a sustainable society, the achievable potential will be significantly constrained. The effect is especially significant in lower solar radiation regions. The European Union holds only 2% of the global solar net energy potential.
... According to Table 2, the CDF plots are shown at two specific projections of 90 and 70%. Statistically sparking, 70% is a reasonable probability selection, since it has been used as a rule of thumb in order to incorporate the data of a CDF model to actual representation of its findings, which is a practice that has been widely utilized [42][43][44]. ...
Article
Full-text available
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http:// creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com Abstract The performance of a photovoltaic (PV) installation is affected by its tilt and azimuth angles, because these parameters change the amount of solar energy absorbed by the surface of the PV modules. Therefore, this paper demonstrates the impact of the azimuth angle on the energy production of PV installations. Two different PV sites were studied, where the first comprises PV systems installed at-13°,-4°, +12° and +21° azimuth angles in different geographical locations, whereas the second PV site included adjacent PV systems installed at-87°,-32°, +2° and +17° azimuth angles. All the investigated PV sites were located in Huddersfield, UK. In summary, the results indicate that PV systems installed between-4° and +2° presented the maximum energy production over the last 4 years, while the worst energy generation were observed for the PV system installed at an azimuth angle of-87°. Finally, the probability projections for all observed azimuth angles datasets have been assessed. Since PV systems are affected by various environmental conditions such as fluctuations in the wind, humidity, solar irradiance and ambient temperature, ultimately, these factors would affect the annual energy generation of the PV installations. For that reason, we have analysed the disparities and the probability of the annual energy production for multiple PV systems installed at different azimuth angles ranging from-90° to +90° degrees, and affected by different environmental conditions. These analyses are based on the cumulative density function modelling technique as well as the normal distribution function.
Article
The measurement of near real-time national solar PV electricity output is an increasingly important component for managing electricity systems. PV systems embedded in the distribution network are problematic for system operators since their output is not directly visible through existing transmission system metering. Typically, national solar PV output is measured indirectly by upscaling the output from a sample of reference PV systems using an estimate of the national capacity. National PV output is then used, along with similar measurements for other embedded technologies such as wind, to train and validate electricity forecasts which ensure efficient electricity market operation. For the first time and using Great Britain as a case study, we investigate the accuracy of this general approach by characterising different sources of uncertainty in national PV output measurements. We find that the capacity error, at ±5%, dominates the yield calculation error, at < ±1% and leads to an overall error in GB solar PV output estimates of ±5.1%. We conclude that solar PV measurements, and consequently national electricity demand forecasts, are currently limited by the state of national PV capacity registers. To make PV output measurements more accurate nations must develop a coherent and comprehensive approach to PV system registration, and output measurement methods must move away from using static solar PV capacity registers towards estimating an operational grid-connected solar PV capacity. This could involve moving away from capacity registers altogether and instead estimating capacity from other sources, for example, from network power flows, satellite imagery, or more likely, a combination of complementary datasets.
Article
Full-text available
This article presents an initial performance analysis of a database of photovoltaic (PV) system performance time series collected within the European funded COST Action PEARL PV. The database contains monitoring data of over 8400 PV systems with accompanying metadata. The PV plants are small residential systems, primarily installed in Europe, with a high density in Belgium. In this initial study, the annual average performance ratio, the annual energy yield, and the performance loss rate of the systems are determined and evaluated. The systems have an average lifetime of 30.5 months. The annual mean performance ratio across all systems is 76.7% and the average yield is 954.9 kWh/kWp per year. The performance loss rate is calculated using three different statistical approaches and one irradiance data source. Average performance losses between $-$ 0.74%/year and $-$ 0.86%/year are calculated depending on the used approach. Furthermore, certain weather-dependent correlations are detected, such as decreasing performance ratio and increasing yield values with increasing irradiation. This study is a stepping-stone for further populating the present database, lessons learnt for handling large amounts of PV performance data, and carrying out performance studies of PV system fleets installed across Europe.
Article
Quantifying small scale domestic solar (PV) generation from energy consumption is becoming increasingly important as the install base of small solar (PV) panels rapidly grows. Unfortunately, it is often the case that the only insight into the consumption and generation of energy within a house comes from smart-meter readings. The smart meter records the amount of energy the house takes from the grid, and does not independently measure and report the local generation that might be consumed by the home, or fed back to the grid. To address this issue, we propose a novel approach to disaggregate PV generation from energy consumption that also infers installed PV capacity. This is done by disaggregating PV generation from censored smart meter readings, and specifically by finding the most likely distribution for the energy consumption and using it to infer the solar generation. We extend this approach to propose the first technique to infer PV capacity without weather data or a solar proxy, using instead only smart meter readings given a group of houses in close proximity. We evaluate the algorithm on two datasets: (i) the US Pecan Street dataset is adapted so that net energy meter readings are censored; and (ii) a constructed dataset, combining smart meter readings from UK households and solar energy generation from locations across the UK. Our results show comparable accuracy at inferring PV capacity compared to existing approaches, which cannot deal with censored readings which represent over 50% of PV panel installations in the UK.
Article
Full-text available
Solar inverter providers optionally offer plant operators online access to their installed photovoltaic (PV) systems. Benefits consist of better system monitoring, faster response to maintenance needs, and a more comfortable data acquisition. Data mining from an online database enables a comparative statistical analysis for a large number of plants about specific productivity parameters, emission reductions and monetary savings. Restriction of information usually confines the progress of PV systems. Due to this situation, the Sunnyportal website managed by one of the most important world provider of solar inverters does not delimit the visualization of their own systems. In addition, there is unrestricted access to their available plants in cases where the owners set the status of their plants as “public”. This study gathered and analyzed information on photovoltaic installations in Mexico with free access to production data. The results show an increase of almost 700% in the number of registered plants in Mexico with exponential growth since 2012. Mexico still shows a lack of studies regarding the monitoring of PV systems because available data are scarce. The specific productivity rate of the PV systems primarily depends on solar irradiation on-site, weather conditions, air contamination, the technology employed, correct engineering and plant maintenance. Based on data obtained, conclusions are drawn concerning preferences and performance of Mexican PV plants. The most both surprising as well as important result is that there is no correlation between specific productivity and average daily on-site irradiation.
Conference Paper
Full-text available
Performance statistics for an ensemble of distributed PV systems in the UK are rare, and there are few publications based on an ensemble representative of the UK PV fleet. This work presents the results of analysing the energy production and the performance of more than 2500 distributed PV systems in the UK monitored for up to 7 years, and provides an overview of the state of the art. We have mapped the annual energy yield of all the PV systems in the database according to the regions of the UK. The mean reported annual energy yield for PV in the UK in 2013 was 910 kWh/kWp, and we assessed the corresponding long-term value as 886 kWh/kWp. The mean value of the yearly-integrated PR for the years 2012-2014 was 84% (with 3% standard deviation). We have shown the presence of seasonal trends in mean monthly PR, with peaks in Spring/Autumn and troughs in Summer/Winter. We have found an increase in the overall performance of PV systems between 6% and 18% from 2004 to 2012, which represents a clear increasing trend with a mean gradient of 1.5%/year.
Conference Paper
Full-text available
Yield predictions are performed to predict the solar resource, the performance and the energy production over the expected life time of a Photovoltaic (PV) system. In this study we compare yield predictions and monitored data for 26 PV power plants. The monitoring data includes in-plane irradiance for comparison with the estimated solar resource and energy yield for comparison with predicted performance. The results show that, due to increased irradiance in recent years (" global brightening ") the yield predictions systematically underestimate the energy yield of PV systems by about 5%. As common irradiance databases and averaging times were used for the yield predictions analyzed in this paper, it is concluded, that yield predictions for areas where the global brightening effect occurred in general underestimated the energy yield by the same magnitude. Using recent satellite derived irradiance time series avoids this underestimation. The observed Performance Ratio of the analyzed systems decreases by 0.5 %/year in average with a relatively high spread between individual systems. This decrease is a main factor for the combined uncertainty of yield predictions. It is attributed to non-reversible degradation of PV cells or modules and to reversible effects, like soiling. Based on the results of the validation the combined uncertainty of state of the art yield predictions using recent solar irradiance data is estimated to about 8%. Keywords: PV systems, yield prediction, performance ratio, solar resource assessment, global dimming and brightening
Conference Paper
Full-text available
Performance statistics for an ensemble of distributed PV systems in the UK are rare, and there are few publications based on an ensemble representative of the UK PV fleet. This work presents the results of analysing the energy production and the performance of more than 2500 distributed PV systems in the UK monitored for up to 7 years, and provides an overview of the state of the art. We have mapped the annual energy yield of all the PV systems in the database according to the regions of the UK. The mean reported annual energy yield for PV in the UK in 2013 was 910 kWh/kWp, and we assessed the corresponding long-term value as 886 kWh/kWp. The mean value of the yearly-integrated PR for the years 2012-2014 was 84% (with 3% standard deviation). We have shown the presence of seasonal trends in mean monthly PR, with peaks in Spring/Autumn and troughs in Summer/Winter. We have found an increase in the overall performance of PV systems between 6% and 18% from 2004 to 2012, which represents a clear increasing trend with a mean gradient of 1.5%/year.
Conference Paper
Full-text available
BIPV systems are small PV generation units spread out over the territory, and whose characteristics are very diverse. This makes difficult a cost-effective procedure for monitoring, fault detection, performance analyses, operation and maintenance. As a result, many problems affecting BIPV systems go undetected. In order to carry out effective automatic fault detection procedures, we need a performance indicator that is reliable and that can be applied on many PV systems at a very low cost. The existing approaches for analyzing the performance of PV systems are often based on the Performance Ratio (PR), whose accuracy depends on good solar irradiation data, which in turn can be very difficult to obtain or cost-prohibitive for the BIPV owner. We present an alternative fault detection procedure based on a performance indicator that can be constructed on the sole basis of the energy production data measured at the BIPV systems. This procedure does not require the input of operating conditions data, such as solar irradiation, air temperature, or wind speed. The performance indicator, called Performance to Peers (P2P), is constructed from spatial and temporal correlations between the energy output of neighboring and similar PV systems. This method was developed from the analysis of the energy production data of approximately 10,000 BIPV systems located in Europe. The results of our procedure are illustrated on the hourly, daily and monthly data monitored during one year at one BIPV system located in the South of Belgium. Our results confirm that it is possible to carry out automatic fault detection procedures without solar irradiation data. P2P proves to be more stable than PR most of the time, and thus constitutes a more reliable performance indicator for fault detection procedures. We also discuss the main limitations of this novel methodology, and we suggest several future lines of research that seem promising to improve on these procedures.
Article
Sources of error in the performance of large ensembles of spatially distributed photovoltaic generators are investigated and reported. Errors are propagated to estimate uncertainty in modeled global tilted radiation and performance ratio (PR)(PR) for the typical UK generator. Uncertainties in generators’ azimuth and elevation lead to typical monthly errors of 4% and 1% on global tilted radiation and PRPR. Interpolation of global horizontal irradiance is affected by an average 5% monthly error and the conversion to the inclined plane leads to an estimated error from 7% to 8% on tilted radiation and PRPR. This prediction has been verified against a set of twenty pyranometers on the plane of the array deployed across the UK, which gauge a 6% monthly error. Mutual cancellations lower this value to 4% for annual periodicity. The estimated monthly error on interpolated global horizontal irradiance is half of the 10% error affecting widely-used Photovoltaic Geographical Information System (PVGIS), which experiences larger errors also on the inclined plane. The assessed uncertainties impact the net present value of the investment required for deploying a PV generator; such an impact has been quantified. The yearly PRPR for the typical UK microgenerator is 84%, a value 8% (6%) higher than recent studies in France (Belgium). In winter, the typical UK performance ratio drops to 75%, because of an increase in shading. Summer performance ratio remains greater than the yearly mean, possibly reflecting the relatively short intervals during which direct sunlight heats the PV modules and the windy conditions over the British Isles. The monthly/annual error affecting the typical individual generator virtually cancel out for the whole national ensemble.
Article
The Photovoltaic Geographical Information System (PVGIS) is a web application for the estimation of the performance of photovoltaic (PV) systems in Europe and Africa, which has become widely used by the PV community in Europe. We here present the results of adapting the solar radiation data calculated from satellite data in the Climate Monitoring Satellite Application Facility (CM-SAF) to PVGIS. The CM-SAF solar radiation database is characterized by very low overall bias and shows good accuracy at validation sites. The application to PVGIS brings important improvements relative to the existing solar radiation databases within PVGIS.
Thesis
This document summarizes the equations and applications associated with the photovoltaic array performance model developed at Sandia National Laboratories over the last twelve years. Electrical, thermal, and optical characteristics for photovoltaic modules are included in the model, and the model is designed to use hourly solar resource and meteorological data. The versatility and accuracy of the model has been validated for flat-plate modules (all technologies) and for concentrator modules, as well as for large arrays of modules. Applications include system design and sizing, 'translation' of field performance measurements to standard reporting conditions, system performance optimization, and real-time comparison of measured versus expected system performance.