Professional Documents
Culture Documents
Quality Assurance of Large Scale PV Power Plants
Quality Assurance of Large Scale PV Power Plants
net/publication/254027313
CITATIONS READS
12 553
4 authors, including:
Christian Reise
Fraunhofer Institute for Solar Energy Systems ISE
65 PUBLICATIONS 1,428 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
PERFORMANCE - A science base on photovoltaics performance for increased market transparency and customer confidence View project
All content following this page was uploaded by Nils Holger Reich on 06 April 2021.
INTRODUCTION
Authorized licensed use limited to: FhI fur Solare Energiesysteme. Downloaded on April 06,2021 at 16:29:00 UTC from IEEE Xplore. Restrictions apply.
Actual system power output is then calculated based upon module. Usually, only about 0.05% to 0.5% of the total
simulated operating conditions, mainly irradiation intensity number of modules is needed for the comparison with
incident onto the module plane and module temperature. power ratings obtained by independent testing facilities.
The determination of module power output at conditions This way, potential deviations in manufacturer-indicated
other than STC is usually based upon model parameters power output claims can be uncovered, enabling more
derived from data sheet specifications. However, a reliable information on the actual power output of all
comparison of parameters derived this way with modules shipped by the particular supplier.
parameters determined by own measurements revealed
relatively high deviations [2]. Interestingly, errors in
simulated yield related to deviating temperature
coefficients reportedly do not have as large an effect as
deviations in low-light characterization of modules. It
therefore seems desirable to at least test for the low-light
performance of modules during module measurements,
which are currently used primarily for testing the rated
STC module power output (see following section).
Regarding irradiation data, row shading and external
shading may need to be considered in addition to deriving
HPOA from global horizontal irradiation data. Modelling the
influence of shading, however, is not easy, because,
amongst other issues, one shaded cell of one single
module may affect entire module strings. An example for
ongoing research in this area is an electrical model that Figure 3: Example of module power output verification
distinguishes each cell for shaded solar generators, with by random sampling approach in module testing.
initial results already having been presented [3].
One example for module power output verification by
MODULE MEASUREMENTS random sampling is given in Fig. 3, showing comparisons
for 20 modules from each of the 4 manufacturers. The
Once planning comes to an end and actual modules are rated module power (as labelled on the back of the
purchased, it is essential to assure that shipped modules modules) and actually measured power output by
actually meet their specifications. For example, for a one- Fraunhofer ISE’s CalLab Modules is depicted in this
megawatt plant and today’s cost- and feed-in-tariff graph. As shown, overrated modules were shipped by
structure, a financial loss of as much as ~50,000€ would manufacturer 2, for which the largest difference of
result over a plant operating period of 20 years, if average manufacturer-claimed vs. ISE-measured efficiency was
module power is only one percent lower. ~4%. However, underrated modules were also detected,
In order to ensure that irregularities are identified before which was likely caused by so-called “plus-tolerances” in
installation, the measurement of a representative number the module purchase contract.
of modules is recommended, preferably in accredited
laboratories. An accredited laboratory ensures that work is SYSTEM TESTING
conducted according to state-of-the-art standards [4-8]
and that measurements with only small uncertainties of To assure that systems already in operation deliver their
reported figures are trustworthy. Prerequisites for the latter specified power output, it is necessary to test the entire
are traceable calibrations of all parts of the measurement system. Tests may include both the identification of
equipment, an uncertainty calculation according to (general) defects and the documentation of deviations in
international standards [9, 10] and a thorough quality system specifications.
management including regular international round robin On-site testing of systems can also contribute to the
tests [11]. Frainhofer ISE’s CalLab PV Modules provides verification of module power, which has actually been
measurement uncertainties of 2% for crystalline silicon installed. The procedure and general test setup is shown
modules and 3.5% to 5% for most thin film modules, in Fig. 4. Here, the total installed PV capacity at a site is
depending on the technology and measurement procedure tested by I-V curve measurements of sub-arrays or
chosen. Temperature coefficients and low light individual strings, usually for ~20% of all modules installed
performance can also be verified in this laboratory. at a site. This may reveal weak module power and/or
Testing a “representative number of modules” for a PV cabling faults. One downside of on-site system testing is
power plant in the MW range, however, is next to the dependency on weather conditions: Field I-V curve
impossible: A number-related definition of “statistically measurements can only be performed under clear skies
representative” will result in an unfeasibly large amount of with fairly high and constant irradiance (> 800 W/m²),
measurements. A special distribution-related procedure to which is due to the need for extrapolation from actual
select representative module batches from manufacturers’ conditions deviating from STC [12]. This also requires
“flash lists” was therefore developed. A “flash list” careful determination of temperature coefficients of the
indicates the manufacturer-measured power of each module (type) in question. Measurement uncertainties of
on-site power verification are roughly 2.5% to 5%,
Authorized licensed use limited to: FhI fur Solare Energiesysteme. Downloaded on April 06,2021 at 16:29:00 UTC from IEEE Xplore. Restrictions apply.
depending on the current weather conditions, assuming request an independent system test with a report that
I-V curves are measured according to standards, using includes a comparison between the as-built and as-
primary calibrated equipment and meticulously determined planned system.
parameters, as presented in [13]. In addition, external
influences such as module soiling or electrical losses need YIELD MONITORING
careful consideration.
Continuous monitoring of PV systems assures the early
detection of system faults (such as inverter breakdowns)
and long-term performance analysis (such as a detailed
I-V curve quantification of loss-mechanisms and annual variations of
Measurement
the PR related to changing meteorological conditions). In
addition, monitoring also helps in validating and further
developing models and yield simulation software.
Measured Power
Data acquisition systems of Fraunhofer ISE monitoring
services typically record the plane-of-array irradiation
HPOA, module temperature, ambient temperature,
CORRECTION TO STC
Power at STC Temperature measured total AC power output by connecting to the
& Irradiance utility feed-in meter, and measured DC and AC power
Measurement
ACCOUNTING FOR output of reference (sub) systems to test for DC and AC
Power at STC, performances in the systems.
corrected for losses Irradiation is measured by calibrated irradiance sensors
Soiling DC wiring
Wiring losses
with uncertainty of ± 2%. For larger systems, often both a
VERIFICATION OF
Electrical mismatch pyranometer and a monocrystalline silicon sensor
(referred to as simply ‘reference cell’) are installed.
Labeled Module Power
Involved measurement uncertainties are listed in the
following section (see Table I).
Figure 4: On-site measurement of solar generator I-V
The DC output of a reference (sub) system is measured
curves for testing installed PV capacity.
between the solar generator and inverter, which allows for
the calculation of solar generator efficiency (SG) and
In addition, system testing may also include more detailed
inverter efficiency (Inv) on top of the Performance Ratio
examinations, such as the identification of damaged
(PR) of the plant as a whole. Monitoring data is acquired
modules or modules not working optimally by thermograph
from 1-second measurements and is usually aggregated
images obtained from infrared cameras, see Fig. 5.
to 5-minute averages.
Exemplary charts from monitored data are shown in Fig. 6,
depicting measured PV efficiencies, currents and voltages
as a function of measured irradiation intensity. Also
inverter efficiency as a function of normalized DC to AC
power output is shown. A detailed analysis of monitoring
data and comparisons of simulated with monitored (sub)
system performances will be given elsewhere [1].
Authorized licensed use limited to: FhI fur Solare Energiesysteme. Downloaded on April 06,2021 at 16:29:00 UTC from IEEE Xplore. Restrictions apply.
ANALYSIS OF LONG-TERM MODULE STABILITY In total, monitoring data used for this study included 489
USING MONITORING DATA years of data, was performed for 107 systems, 16 of which
have modules with either string-ribbon or edge-defined
Long-term PV module stability is usually investigated by film-fed growth (EFG) solar cells and all others having
focusing upon individual modules installed outdoors over standard c-Si cells. We excluded systems with thin film
long time periods, or by accelerated lifetime testing modules in the analysis presented here. On average, data
accompanied by subsequent or continuous module availability was 99.4% for all systems in the entire period.
characterization in the laboratory. Both these approaches,
however, are cumbersome in that modules need to be Resulting “rates of change” of module power
handled and measured individually. In addition, results are
not necessarily representative, because only individual Obtained change rates from above described filtering
circumstances of these individual modules are focused on, procedure varied between –3%/yr…+3%/yr for standard c-
in regards to both the conditions the modules are operated Si modules in the initial ~5-7 years of operation. The
at and the modules themselves. To this end, we embarked statistical spread in changing power output over time is
upon an analysis of PV module degradation by means of evenly distributed, i.e., about the same decrease in PV
evaluating a large quantity of monitoring data of entire PV power output (usually referred to as degradation) as well
power plants [14, 15]. as an increase was observed. For EFG and string ribbon
In order to exclude influences from different climatic solar cells, however, the resulting rate of change is
conditions in different years, data points at comparable consistently negative. The results are illustrated in Fig 9
and failure-free conditions had to be selected. We decided for all individual systems.
on the following steps:
- Filtering of 5-minute averages at irradiation intensity
between 800 W/m² and 1000 W/m²:
- Avoiding shading and angle of incidence effects
due to relatively high sun elevation
- Efficiency of crystalline silicon based modules is
almost constant between 800-1000 W/m²
- Binning data points by temperature into 5K intervals:
The bin with the most data points was selected for the
analysis of each system, typically 40°C…45°C
- Exclusion of outliers: All data points with a deviation
of more than ± 5% from the annual median
Around 700 to 1000 data points remained per year for the
analyzed systems, for which a linear slope was fitted to
derive an annual rate of change. The filtering procedure is
illustrated in Fig. 7.
Figure 8: The ‘rate of change’ of rated power output
estimates (in %/yr) resulting from the degradation
analysis using monitoring data.
Authorized licensed use limited to: FhI fur Solare Energiesysteme. Downloaded on April 06,2021 at 16:29:00 UTC from IEEE Xplore. Restrictions apply.
monitor crystalline silicon modules, effects due to However, more detailed studies are needed before we are
changing spectral conditions supposedly level out too. The able to include uncertainty figures related to soiling effects.
most critical quantity in the monitoring-measurements is The estimated combined uncertainty of monitoring data is
irradiance intensity, notably the relatively large initial detailed in Table I. As listed, we anticipate ~3%...4% in
uncertainty of ~2% in reference cell calibration. In addition, overall uncertainty, which is about the same as the initial
uncertainty introduced by temperature correction of the variation of rates of change. This variation of the rate of
reference cell signal has to be considered [13]. change, for all systems, is illustrated in Fig. 10, which
shows the distribution of calculated rates of change for all
Global spectral irrandiance systems monitored >1.5 years and >5 years.
on Module + sensor plane
Calibration value
Module output
Stability of performance
Rate of change
Authorized licensed use limited to: FhI fur Solare Energiesysteme. Downloaded on April 06,2021 at 16:29:00 UTC from IEEE Xplore. Restrictions apply.
yield monitoring. Especially continuous monitoring of PV [5] IEC60904-2 Ed.2 (2007-03), Photovoltaic devices –
systems is relevant, as it assures the early detection of Part 2: Requirements for reference solar devices
system faults (such as inverter breakdowns). In addition,
monitoring also helps in validating and further developing [6] IEC60904-4 Ed.1 (2009-06), Photovoltaic devices –
yield simulation software and allows for long-term Part 4: Reference solar devices – Procedures for
performance analyses. Combined, all quality assurance establishing calibration Traceability
measures ascertain that high performances ratio of up to
90% can not only be achieved but also sustained. [7] IEC60904-7 Ed.3 (2008-11), Photovoltaic devices –
One issue of sustained high PV performance is so-called Part 7: Computation of the spectral mismatch correction
module degradation. Most often, degradation rates are for measurements of photovoltaic devices
investigated by focusing upon individual modules, either
installed outdoors during long time periods or in the [8] IEC60904-9 Ed.2 (2007-10), Photovoltaic devices –
laboratory using accelerated lifetime testing. Both these Part 9: Solar simulator performance requirements
approaches, however, are cumbersome in that modules
need to be handled and measured individually. In addition, [9] ISO/IEC Guide 98-3:2008: Uncertainty of measurement
results are not necessarily representative, because of the – Part 3: Guide to the expression of uncertainty in
focus on only individual circumstances. To this end, we measurement
presented an analysis of PV module degradation by
means of evaluating a large quantity of monitoring data of [10] Müllejans et al.,”Analysis and mitigation of
entire PV power plants. measurement uncertainties in the traceability chain fort he
Rates of change were determined in an analysis of calibration of photovoltaic devices”, Measuring Science
module degradation using monitored system performance and Technology 2009 (20) pp. 1-6.
data of 107 systems including 489 years of monitoring
experience. Obtained rates of change varied [11] Werner Herrmann, Klaus Kiefer et al.,”PV Module
between -3%/yr…+3%/yr for standard c-Si modules in the output power characterisation in test laboratories and in
initial ~5-7 years of operation. For EFG and string ribbon the PV industry – Results of the European Performance
solar cells, however, the obtained rate of change was Project“, 25th EU PVSEC, Valencia, Spain, pp. 3879 -
consistently negative, suggesting it is a feasible approach 3883
to use monitoring data of large-scale systems to perform
power degradation analyses. [12] IEC60891 Ed. 2 (2009-12), Photovoltaic devices -
Statistical spread in changing power over time of standard Procedures for temperature and irradiance corrections to
c-Si modules was evenly distributed. Such a statistical measured I-V characteristics
spread with an average close to zero leads to the
conclusion that it is not necessarily relevant to consider [13] D. Dirnberger et al.,”Uncertainty of field I-V-curve
degradation of standard crystalline silicon modules on the Measurements in large Scale PV-Systems“, 25th EU
system level in the initial years of plant operation. PVSEC, Valencia, Spain, 2010, pp. 4587 - 4594
In conclusion, today’s high Performance Ratio prove solar
PV has matured into both a highly reliable and well- [14] K. Kiefer et al., “A Degradation Analysis of PV Power
performing technology. On top of increased component Plants”, 25th EUPVSEC, 2010, Valencia, Spain, pp. 5032
efficiencies, also quality assurance measures contribute in – 5037
achieving and sustaining these high system performances.
[15] M. Jantsch and Ch. Reise,”PV module and system
REFERENCES degradation including BOS effects”, Report D4.5.1 of the
“Performance” project 2010, www.pv-performance.org
[1] N.H. Reich et al.,”Performance Ratio revisited: Are
th
PR > 90% realistic”, to be presented at the 26 EUPVSEC,
Hamburg, Germany, 2011
Authorized licensed use limited to: FhI fur Solare Energiesysteme. Downloaded on April 06,2021 at 16:29:00 UTC from IEEE Xplore. Restrictions apply.
View publication stats