Performance Parameters for Grid-Connected PV Systems

B. Marion,’ J. Adelstein,’ K. 8oyle.l H. Hayden.’ B. Hammond,’ T. Fletcher,’

B. Canada,’ D. Narang,‘ A. Kimber.3 L. Mit~hell.~ G. Rich,4 and T. fownsend4

Najional Renewable Energy Laboratory, 3617 Cole Blvd., Golden, CO 80401

Arizona Public Service Co., 1500 E. University Dr., Tempe, AZ 85281
3PowerLightCorporation, 2954 San Pablo Ave., Berkeley. CA 94702
First Solar, 4050 E. Cotton Center Blvd. #6-68,Phoenix. AZ 85040

ABSTRACT credibility for the PV industry and positioning it for further

growth.
The use of appropriate performance parameters
facilitates the comparison of grid-connected photovoltaic Parameters describing energy quantities for the PV
(PV) systems that may differ with respect to design, system and its components have been established by the
technology, or geographic location. Four performance International Energy Agency (IEA) Photovoltaic Power
parameters that define the overall system performance Systems Program and are described in the IEC standard
with respect to the energy production, solar resource, and 61724 [I].(IEA task members have used these
overall effect of system losses are the following: final PV performance parameters to develop a database of
system yield, reference yield, Performance ratio, and operational and reliability performance [2]. The database
PVUSA rating. These performance parameters are contains information for several hundred PV systems and
discussed for their suitability in providing desired may be viewed at www.task2.org.)
information for PV system design and performance
evaluation and are demonstrated for a variety of, Three of the lEC standard 61724 performance
technologies, designs, and geographic locations. Also parameters may be used to define the overall system
discussed are methodologies for determining system a s . performance with respect to the energy production, solar
power ratings in the design phase using multipliers resource, and overafl effect of system losses. These
developed from measured performance parameters. parameters are the final PV system yield, reference yield,
and performance ratio.
INTRODUCTION
The final PV system yield YFis the net energy output
Accurate and consistent evaluations of photovoltaic E divided by the nameplate d.c. power POof the installed
(PV) system performance are critical for the continuing PV array. It represents the number of hours that the PV
development of the PV industry. For component array would need to operate at its rated power to provide
manufacturers. performance evaluations are benchmarks the same energy. The units are hours or kWhlkW, with the
of quality for existing products. For research and fatter preferred by the authors because it describes the
development teams, they are a key metric for helping to quantities used to derive the parameter. The Yf normalizes
identify future needs. For systems integrators and end the energy produced with respect to the system size;
customers, they are vital tools for evaluating products and consequently, it is a convenient way to compare the
product quality to guide future decision-making. energy produced by PV systems of differing size:

As the industry has grown, a clear need has arisen for E

greater use of and education about appropriate industry- Y, =- (kWhlkW) or (hours)
standard performance parameters for PV systems. These Po
performance parameters allow the detectionof operational
problems; facilitate the comparison of systems that may The reference yield Y,is the total in-plane irradiance
differ with respect to design, technology, or geographic H divided by the Pv’s reference irradiance G. It represents
location; and validate models for system performance an equivalent number of hours at the reference irradiance.
estimation during the design phase. Industry-wide use of If G equals 1 kW/mz, then Y , is the number of peak sun-
standard performance parameters and system ratings will hours or the solar radiation in units of kWhlm’. The Y,
assist investors in evaluating different proposals and defines the solar radiation resource for the PV system. It is
technologies, giving them greater confidence in their own a function of the focation, orientation of the PV array, and
ability to procure and maintain reliable, high-quality month-to-monthand year-to-year weather variability:
systems. Standard methods of evaluation and rating will
also help to set appropriate expectations for performance H
y =- (hours)
with educated customers, ultimately leading to increased G

0-7803 -8 707405/$20.0002005 IEEE. 1601

 The performance ratio PR is the Yt divided by the Yr. respect to an a.c. power rating, two systems might report
By normalizing with respect to irradiance, it quantifies the the same Yf, but have significantly different inverter
overall effect of losses on the rated output due to: inverter efficiencies, or other loss mechanisms. Similarly, if
inefficiency, and wiring, mismatch, and other losses when performance was with respect to PTC, two systems might
converting from d.c. to a.c. power; PV module report the same Yf,but have significantly different PV
temperature; incomplete use of irradiance by reflection module temperature-related losses because of how the
from the module front surface; soiling or snow: system PV modules are mounted or integrated into the building.
down-time; and component failures:
Although a nameplate d.c. power rating is used in Yf
to report the normalized energy produced by an existing
PR = -y, (dimensionless) (3) system, an ax. power rating is essential when attempting
Y to predict the energy a PV system will produce using
models such as PVWATTS [5], PVDesignPro [6], or
PR values are typically reported on a monthly or PVGRID [7]. Accurate energy predictions are crucial to the
yearly basis. Values calculated for smaller intervals, such continued development of the photovoltaic industry
as weekly or daily, may be useful for identifying because they set the investor's expectations for system
occurrences of component failures. Because of losses due performance and the associated economic retum. The
to PV module temperature. PR values are greater in the remainder of this section discusses a s . power ratings and
winter than in the summer and normally fall within the considerations in their determination.
range of 0.6 to 0.8. If PV module soiling is seasonal, it
may also impact differences in PR from summer to winter. PV systems may be assigned a.c. power ratings by
Decreasing yearly values may indicate a permanent loss accounting for: (1) losses in converting from d.c. to a x .
in performance. power, and (2) operating cell temperatures that are usually
greater than 25°C.In the'first case, the nameplate d.c.
The PVUSA rating method [3] uses a regression power rating is multiplied by empirically determined derate
model and system performance and meteorological data factors to calculate an a.c. power rating at STC. In the
to calculate power at PVUSA Test Conditions (PTC), second case, an additional derate can be applied for
where PTC are defined as 1000 Wlm2 planeof-array temperature other than STC. Finally, the PVUSA rating
irradiance, 20°C ambient temperature, and 1 mls wind method may be used to assign an a.c. rating to an existing
speed. PTC differs from standard test conditions (STC) in system with historical data.
that its test conditions of ambient temperature and wind
speed will result in a cell temperature of about 50"C, To evaluate the accuracy of our empirical derate
instead of the 25°C for STC. This is for a rack-mounted PV factors, PVUSA ratings were determined for 24
module with relatively good cooling on both sides of the PowerLight PV systems (twenty single-crystalline silicon,
module. For PV modules mounted close to the roof or two multicrystalline silicon, and two amorphous silicon)
integrated into the building with the airtlow restricted, PTC located throughout the United States. These ratings were
will yield greater cell temperatures. Nordmann and then compared to the a x . ratings for the same systems
Clavadetscher[4] report that PV module temperatures rise calculated by using the derate method and the derate
above ambient for fielded system ranging from 20°C to factors from Table I.All derate factors in Table 1 were
52°C at 1000 W/m2. with the largest temperature rise for estimated from measured losses and component
an integrated faqade. The difference between the specifications. The typical overall derate factor at nominal
nameplate d.c. power rating and the system PVUSA rating operating cell temperature (NOCT) is 0.731, representing
is an indication of the total system losses associated with a loss of 26.9% from the nameplate d.c. rating.
converfing d.c. module energy to ax. energy. As with
decreasing PR values, decreasing PVUSA ratings over Table 1. Derate Factors for A.C. Power Rating
time may indicate a permanent loss in performance. Item Typical Range
PV module nameplate d.c. rating 1 .oo 0.85- 1.05
D.C. AND A.C. RATINGS Initial light-induced degradation 0.98 -
0.90 0.99
d.c. cabling 0.98 -
0.97 0.99
The Yt is calculated by dividing the energy yield Diodes and connections 0.995 0.99 - 0.997
recorded with a utility kWh meter by the nameplate d.c. Mismatch 0.98 0.97- 0.995
power rating. The nameplate d.c. power rating is Power-conditioning unit (inverter) 0.96 -
0.93 0.97
determined by summing the module powers listed on the
Transformers 0.97 -
0.96 0.99
ax. wiring 0.99 0.98- 0.993
nameplates on the backsides of the individual PV modules Soiling 0.95 0.75- 0.995
in the PV amy. The PV module power ratings are for STC Shading** 1 .oo 0.0 - 1.oo
of 1OD0 Wlm solar irradiance and 25°C cell temperature. Sun-tracking 1.oo -
0.98 1.oo
Besides being easily determined, the nameplate d.c. Availability of system 0.98 -
0.0 0.995
power rating's use in the equation for Yr offers the Overall at STC 0.804 0.62"- 0.964
advantage. as compared to the use of an ax. power rating Temperature (NOCT = 45"C, power 0.91
or conditions other than STC. of differentiating between correction factor = -0.45%/"C)
systems with different d.c. to a x . conversion efficiencies Overall at NOCT 0.731
or different mounting-related PV module temperature *Does not indude soiting. shading, tracking, or availability losses
Typically 0.975for fixdtilt rack-mounted systems
environments. For example, if performance was with

1602
 For the initial comparison, all 'typical" derate factors Consequently, for the second comparison, results
from Table 1 were used, except for the temperature derate were significantly improved by using a derate factor to
factors that were determined using the manufacturers' account for the accuracy of the manufacturer's nameplate
power correction factors for temperature and NOCTs of d.c. ratings, as detailed in the first row of Table I.
45°C. The results of the comparison using this derate Compared to the PVWSA ratings, the a.c ratings
method are shown in Fig. 1 for a representative sample of calculated using a derate method including a factor for
the systems evaluated. For all systems, the derate manufacturer's nameplate rating were within *5%, with a
method ax. ratings were as much as 19% greater than the standard deviation of the differences of 2%. Figure 2
PVUSA rating, and the standard deviation of the illustrates these results. Although not evaluated, still better
differences was 7%. In Fig. f , the measured loss is the agreement might have been achieved by using systsm-
difference between the nameplate d.c. rating and the specific derate and NOCT values instead of typical values.
PVUSA rating. The design loss is the difference between
the nameplate d.c. rating and the a.c. rating calculated INFLUENCE OF WEATHER
using the derate method. For an accurate design, the
measured loss and design loss will be very close. The Variations in solar radiation and ambient temperature
measured and design losses are epressed as from month-to-month and year-to-year influence the
percentages for ease of comparison. performance parameters. Therefore, it is important to
identify which performance parameters are suitable for
5o which system evaluations based on their weather-
I .:?~*%?hb&'%?&~ M-uM Loss
Deslgn Loss
dependence. The Yf is influenced the most because of its
-1
4
40 - dependency on solar radiation. The PR is influenced Jess
because values are normalized with respect to solar
radiation, but values are influenced by seesonal variations
t4 30 in temperature. The PVUSA a.c power ratings at PTC are
4
E influenced the least because the method performs the
J 20 regression using solar radiation, ambient temperature. and
i? wind speed values. Small variations in PVUSA method
10 a.c power ratings can be attributed to the range of values
over which the regression is performed, nonlinearities in
PV module and inverter performance, and variations in
0
Awa@ Manl Man2 Man3 Man4 Man% Man%
solar spectrum.

Fig. 1. Design and measured losses using typical derate factors, To illustrate the extent to which the performance
except for the temperature derate factor, which was manufacturer parameters might be influenced by weather, PV system
specific. performance was modeled using PVFORM [9]for a 30-
year period. The howy solar radiation and meteorological
Current-voltage (l-V) curve testing of PV modules data input to PVFORM was for the Boulder, CO, station in
used in these 24 systems revealed that the accuracy of the National Solar Radiation Data Base [+IO]. PV system
the nameplate ratings vaned by manufacturer, and for specificationswere the same as the PV system located on
certain manufacturers the accuracy varied by product. the roof of the Solar Energy Research Facility (SERF) at
Some PV modules produced as much as 4% more than the National Renewable Energy Laboratory (NREL):
specified. whereas others were as much as 12% less than single-crystalline silicon PV modules. nameplate d.c.
specified [8]. power rating of 7420 W. PV array tilt angle of 45", and PV
array azimuth angle of 22" east of south. Using modeled,
50 instead of measured, data permitted the influence of
weather to be evaluated over a longer period of time and
40 eliminated the need to carefully screen erroneous data or
data collected when the system was inoperative, or to
account for any performance degradation that occurred.

Using the modeled data for the 3ayear period,

monthly and yearly performance parameters and WUSA
a.c power ratings at PTC were calculated. The results are
shown in Fig. 3. If weather had no influence, the values
would all reside on a horizontal line, but that is clearly not
the case. As expected, Yf shows the greatest variability
Average M a n l Man2 Man3 Man4 Man% Maneb and the PVUSA a.c power rating at PTC shows the feast.
Although not shown, the variability of Y, is similar to Yf
Fig. 2. Design and measured losses using typical derate factors, because of Yis dependence on solar irradiance. PR
except for the temperature derate factor and the PV module values exhibit the influence of temperature. with smaller
nameplate d.c. rating derate factor, which were manufacturer values in summer than winter. For yearly values. 95%
Specific.
confidence intervals, determined as twice the standard

1603
deviations, are shown. The confidence interval of i8.4% Table 2. Arizona Public Service PV Systems and Their Yf for
for Yf means that 95% of the yearly values should be September 2003 Through August 2004.
within 8.4% of the average yearly value. As indicated by System Name Location Size Yf
the scatter of data, monthly values are more variable, (kWk) (kWhkW)
Singlehis Tracking. North-South Horizontal Axis
resutting in greater confidence intervals than for the yearly
Embry Riddle PESCOtt 228.50 1906
values. Although PR vanes from summer to winter, the Gilbert Nature Ctr. Tempe/Mesa 244.00 1682
yearly values are consistent with a confidence interval of Ocotillo 1 Tempe 94.47 1806
i1.2%, which is only slightly greater than the confidence Airport MTA2 Prescott 121.oo 2118
interval of *0.7% for yearly values of PVUSA a.c power Airport MTA7 Prescott 151.20 1882
ratings at PTC. (Because the PVUSA ratings are Airport MTBI PP3scOtt 151.20 1406
determined using a month of data, the yearly value was Airport MTB2 PP3scott 151.20 1807
determined as the average of the 12 monthly values.) Airport MTBS Prescotl 15120 1861
Airport MTBS prescott 117.60 1869
Consequently, both PVUSA a.c power ratings at PTC and
Water Tanks-East Scottsdale . 153.60 1986
yearly PR values should be able to detect degradation of Water Tanks-West Scottsdale 144.00 2020
system performance over time. Yucca Pwr. Plant Yuma 121.oo 2147
Non-Tracking,Horizontal
Star Parking Tempe 5.04 1345
Non-Tracking,South-Facing with Tilt Equal to Latitude
Challenger Peoria 2.28 1593
Desert Lake Pleasant 2.28 1461

0.9 ' 95% Confdenso Interval fw Yearly Valuer =Average *8.4%

Airport MTBI system, the inverter operated poorly until
August, when all its performance issues had been
resolved. The Gilbert Nature Center system experienced
frequent inverter faults, and in August a conductor failed,
rendering the system inoperable or operating at reduced
power for most of the month.

The Yt normalizes performance with respect to

system size; consequently, it is useful for comparing
systems of different size to quantify benefits of design,
components, or locations. But unlike the PR, the Vt values
5000 do not correct for the variability of solar radiation, and
-a
2 therefore, are not as useful for identifying operational
;S 4000 - '
problems. The exception is side-by-side operation of

'
'3

3000
95% Cmfdence Intervalfor Yearly Values = Average i0.m
I
J
l l l
F M A M J
l l l t l
J A S O N DYr
l l l l
systems of identical design, such as the Arizona Public
Service single-axis trackers at the Prescott airport. For this
situation, we can assume that all systems have essentialfy
the same solar resource (Yr), and that any operational
problem may be detected by comparing a system'sYf
Fig. 3.. Monthly and yearly Yt, PR, and PVUSA a.c power rating at
PTC +om FV performance data mdeled ow a 3C~yearperiod
400 4000
show the influence of weather variability. + Embry Riddle I Airport MTB2
0 Girberi Naluie Ctr X Airport MT33
D Ocotillo 1 e kimorl MTBS
EXAMPLE RESULTS FOR Yt
I-
~::
Alrpoa MTA2 7 Witer-Tanks East
A ~ r p o rMTA7
l Water-Tanks West
r Alrporl MTBt 4 Yucca Pm.Plant 3000
S
Arizona Public Service Co. operates numerous grid- - E
connected PV systems within its service territory [ I l l . .s
Table 2 contains a listing of some of these systems and 3
-2000 y,
their Yf values for the Wmonth period of September 2003
>-'
through August 2004. Because the solar r e m u m (Yr) is - f
greater for the single-axis tracking systems, their Yf values >
are larger than those for the non-tracking systems. - 1000
For the single-axis tracking systems, Fig. 4 shows
monthly and yearly Yf values for the 12-month period. Yf
0 I I I 1 I 181 I I I I I I 0
values were largest for the Airport MTA2 and the Yucca
power plant systems, primarily because their PV modules' S O N D J F M A M J J AYr
performance met their nameplate expectations. The other 2003 2004
systems performed at a lower level because of a
combination of factors: PV module performance, inverter Fig. 4. Monthly and yearly Vi for Arizona Public Service single
efficiency, and operational problems. Specifically, for the axis trackers for September 2003 through August 2004.

1604
against that of the other systems. For a single system, a reporting snowfall and for three days in February when the
similar strategy might be used by dividing it into two or system was off. Depending on the amount of snow, daily
more subsystems, with each having their own inverter and PR values as low as zero occurred. The influence of snow
as. metering. is also evident in the weekly and monthly PR values, but
to a lesser extent.
EXAMPLE RESULTS FOR PR
As an example of using PR to measure long-term
The PR is a dimensionless quantity that indicates the changes in performance, Fig. 6 presents-for three PV
overall effect of losses on the rated output. By itself, it systems-the linear least-squarefits of monthly PR values
does not represent the amount of energy produced, over a period of several years. For comparison, results
because a system with a low PR in a high solar resource using the PVUSA method are also shown. Both methods
location might produce more energy than a system with a show similar degradation rates, even though they use
high PR in a low solar resource location. However, for any somewhat different input data. Whereas the calculation of
given system, location, and time; if a change in component PR uses all values of irradiance, the !VUSA method
or design increases the PR, the Yf increases accordingly. restricts irradiance values to 800 Wlm or above. To
PR values are useful for determining if the system is examine only the effects of long-term performance
operating as expected and for identifying the occurrence changes, both methods excluded data when the as.
of problems due to inverter operation (faults/failures, peak- power value indicated the system was not operating. If
power tracking, softwarelcontrol), circuit-breaker trips. instead the intent had been to evaluate overall system
solder-bond failures inside PV module junction boxes, performance, data would not have been excluded and
diode failures, inoperativetrackers, shading, snow, soiling, values would have been less. The results depicted in Fig.
long-term PV system degradation, or other failures. Large 6 are an example of using PR to measure performance
decreases in PA indicate events that significantly impact changes over time, and are not meant as a definitive
performance, such as inverters not operating or circuit- analysis of a PV technology's long-term performance for
breaker trips. Small or moderate decreases in PR indicate Denver or any other location. The relative performance of
that a less severe problem exists. The PR can identify the the three systems was influenced by using inverters that
existence of a problem, but not the cause. The cause of have different operating characteristics and conversion
the problem requires further investigation, which may
include a site visit by maintenance personnel. Decreases 1.0 I , 1400
a-Wa-Sila-S! Ge PV System 1224 W,
in PR from soiling or longterm PV system degradation a
may not be readily evident unless viewed over months, or 1200 5
years in the case of the fatter. Decreases from soiling are E?
P
site- and weatherdependent, with greater soiling (up to
06
1000 2
25% for some California locations) for high-traffic. high-
pollution areas with infrequent rain. 05 800
22
0.4 t PR degradation = 1.5% pRr year 6
For 2001, Fig. 5 presents daily, weekly, and monthly A P V ~ J ~ degradaiion
A. = 1.I % per year
PR values for the NREL SERF PV system described in a
CdSiCdTe W System, 1200 W, e
previous section. For most of the year, the PR values are
consistent with those modeled for the same system and 1200 g
shown in Fig. 1. But for winter and spring months, PR F
values are lower for days coinciding with logbook entries 1000 3

i n $
800 7
1 + PR. degradatim = 1 2% per year
A M U ,degradaboo = 1.4% per year
6
0.8 8000
Singlecrystal si WSystem, 7420 W
,
0.9 E

-
0.6

knhly

bity 1994 1998 2002 2006

0.0 Fig. 6. Long-term degradation rates for three PV systems at
J F M A M J J A S O N D NREL from monthly d u e s of PR and PVUSA ratings. Upper
regression linea from monthly PR values shown by + symbols.
Fig. 5. Daily, weekly, End monthly PR values for the NREL SERF Lower regression lines from monthly PVUSA values shown by 13
PV system for 2001. symbols.

1605
efficiencies. Also, the reliability of these small systems [4JT. Nordmann and L. Clavadetscher, 'Understanding
may not be representative of that of larger systems, and Temperature Effects on PV System Performance",
performance changes may have been different if tested in Proceedings of the 3" World Conference on P V Energy
a different climate or location. For the system using the + Conversion,Osaka, Japan, 2003.
Sila-Sila-Si:Ge PV modules, data collection began after
being deployed for several months and their initial [5]B. Marion and M. Anderberg. "PVWATTS-An Online
performance degradation had occurred. Performance Calculator for Grid-Connected PV Systems."
Proceedings of the ASES Solar 2000 Conference, June
SUMMARY AND FUTURE WORK 16-21, 2000, Madison, W1.

Four performance parameters may be used to define 161 "PVDesign Pro User's Manual," Maui Solar Software
the performance of gridconnected PV systems: final PV Corporation, 1988.
system yield Yt, reference yield Y, performance ratio PR.
and PVUSA rating. The Yf and PR are determined using l'i[ H.Wenger, "PVGrid User's Manud," 1990.
the nameplate d.c. power rating. The Yr is the primary
measure of performance and is expressed in units of [El] A. Detrick and A. Kimber, L. Mitchell, "Performance
kWhlkW. It provides a relative measure of the energy Evaluation Standards for Photovoltaic Modules &
produced and permits comparisons of PV systems of Systems," Proceedings of the 31* EEE fhotovolfaic
different size, design, or technology. If comparisons are Specialists Conference, Lake Buena Vista, FL, 2005 (in
made for different time periods or locations, it should be Press)
recognized that year-to-year variations in the solar
resource will influence Yt. The PR factors out solar 191 D. Menicucci and J. Fernandez. "User's Manual for
resource variations by dividing Yf by the solar radiation PVFORM: A Photovoltaic System Simulation Program for
resource, Yr. This provides a dimensionless quantity that Stand-Alone and Grid-Interactive Applications", SAND85
indicates the overall effect of losses and may be used to 0376, Albuquerque, NM: Sandia National Laboratories,
identify when operational problems occur or to evaluate I988.
long-term changes in performance. As part of an
operational and maintenance program, the PR may be [IO] NSRDB Vol. 1, "User's ManuaMational Solar
used to identify the existence of performance issues. Radiation Data Base (1961-1990)," Golden, CO: National
Renewable Energy Laboratory, 1992.
To further encourage the use of common reporting
and design practices for PV systems, future activities [IllL. Moore, H. Post, H. Hayden, S. Canada and D.
should include: (1) additional work to gain support for an Narang, "Photovoltaic Power Plant Experience at Arizona
industty-standard set of performance parameters and Public Service: A 5-Year Assessment," Progress in
system derating factors, (2) additional measurements for fhofovoltaics: Research and Applications 2005; (in press).
verifying individual derate factors (e.g.. inverter,
transformer, wiring, soiling,). Although using an overall
derate factor yielded ratings close to that of the PVUSA
method, a better knowledge of the individual derate factors
would provide closer agreement and identify areas to
improve system performance, and (3) development of a
'Buyer's Guide" to explain performance parameters and
system rating factors to potential investors and describe
which system aspects are the biggest drivers of
performance (e.g.. inverter efficiency, module efficiency,
reliability, performance degradation rate, system location).

REFERENCES

[l]IEC, 'Photovoltaic System Performance Monitoring-

Guidelines for Measurement, Data Exchange, and
Analysis, IEC Standard 61724," Geneva, Switzerland.
1998.

121 U. Jahn and W. Nasse, "Performance Analysis and

Reliability of Grid-Connected PV Systems in IEA
Countries", Proceedings of the 3"j World Conferenceon
PV Energy Conversion,Osaka, Japan, 2003.

f3] R. Dows, 'PVUSA Procurement, Acceptance, and

Rating Practices for Photovoltaic Power Plants," PG&f
Co. Report H5-30910000.1, Sept. 1995.

1606