1. Outdoor PV performance testing provides key insights but sum kWh/kWp values alone are not enough to understand results.
2. Detailed analysis of DC module performance helps explain AC array data by accounting for factors like losses, mismatch effects, and weather impacts.
3. Techniques like normalizing voltage and current values, examining maximum power over time, and comparing to models enable identification of issues like shading, degradation, and temperature effects that influence energy yields.
1. Outdoor testing, analysis and
performance predictions of PV
technologies
Steve Ransome (Owner SRCL)
and associate consultant (Intertechpira UK)
www.steveransome.com
2. How kWh/kWp values are used
by industry sector
Section of PV industry kWh/kWp relevance
Manufacturers Claim high performance
Indoor testers Measure relevant parameters
Sizing programs Claim accurate predictions from
(simulation models) complex models
Customers Expect high values
Financial backers Demand guaranteed values
over lifetime
Independent outdoor Different rankings for each
comparisons technology (often within
experimental error for correctly
rated and measured modules)
14-May-09 www.steveransome.com Page 2
3. Typical daily AC performance of a
large array, USA
1. Energy Yield YF
(kWh/kWp/d) should be
approximately proportional
to the daily insolation YR
(kWh/m²/d)
2. Points below the line
indicate underperforming
periods
3. The total uncorrected
energy yield will include
these bad points and
worsen apparent yield
14-May-09 www.steveransome.com Page 3
4. Energy yield losses from AC arrays
How much of these energy losses are due to
– component number and choice
– down time
– inverter loss (efficiency or low light turn on)
– inherent differences between module technologies ?
– other reasons ?
DC module performance must be studied to quantify
the losses
14-May-09 www.steveransome.com Page 4
5. View of typical DC setup
ISET, Kassel Germany
“Spectrally
30° sensitive”
Tilt Irradiance
sensors
South PV modules
Not shown :
temperature
sensors
Direct:Diffuse,
precipitation ,
wind speed etc.
Pyranometer
6. Typical DC outdoor measuring setup
(single devices are better for characterisation)
Windspeed (ms-1)
Plane of array Irradiance
(kW/m²)
Impp, Vmpp
Device or IV scan
under
test
Data Logger
Device Temperature (C) Measure
every 1-10
Ambient Temperature (C) minutes
Other sensors ? e.g.
horizontal irradiance,
precipitation, air pressure, Data
spectrum … Analysis
7. Independent energy yield test :
7 technologies Kassel, Germany
1. Most technologies give
similar energy yields
(<±4% kWh/kWp)
2. Two are much lower (are
they faulty or have they
degraded?)
3. Cannot identify reasons
from kWh/kWp sums alone
(see my paper PVSEC
Valencia 2008 for details)
14-May-09 www.steveransome.com Page 7
8. How do we find the reason
for differences in kWh/kWp ?
Possible reasons
• Monitoring errors e.g. Vmax mistracking
• Pmax declaration (measured/nameplate)
• Shading on some panels only
• Degradation/annealing
• Different technology performances at
– low light
– high temperature
– diffuse light
– different spectra …
Detailed studies should reveal reasons for differences
14-May-09 www.steveransome.com Page 8
9. Data validation for
outdoor measurements
Normalise measurements to “measured/expected values”
• Vdm = Vdc / Vmax.stc
• Idn = Idc / Imax.stc / Irradiance
Define simple limits to remove “bad” data points
(e.g. 80-110% of expected value )
Perform a sanity check on meteorological data
Irradiance (e.g. 0 to 1.4kW/m²),
Clearness Index (e.g. 0.2 to 0.8)
Diffuse Fraction (e.g. 0.1 to 0.9)
Temperatures (e.g. -20 < Ambient < 40)
etc.
14-May-09 www.steveransome.com Page 9
10. Normalised electrical parameters
showing limits used for Imax and Vmax
Weather data (top)
Electrical data (bottom)
• Correct, interpolate or delete
data outside sensible limits
(shown in coloured bands)
• “Redundant data” : calculate
“NOCT” (Tmodule
@800W/m², Tambient=20C,
1ms-1 wind) – should be ~47C
14-May-09 www.steveransome.com Page 10
11. Diffuse sky (left) vs Clear sky (right)
affects PV performance
Large attenuation of Beam Little attenuation of Beam
High reflection off clouds Little reflection off clouds
Variable spectrum Spectrum ~ Air Mass
14-May-09 www.steveransome.com Page 11
12. Understanding Efficiency vs Irradiance
Imax and Vmax vs. Diffuse:Beam - cSi
Diffuse Clear
Vmax vs Irradiance Imax vs Irradiance
(2) Diffuse
(1) Error
(2) AOI
Efficiency =
Vmax * Imax
1. Most points should be within narrow limits, outlier data due to poor
tracking, shade or snow on module or sensor. Can temperature correct.
2. Imax differs whether diffuse or clear sky, Vmax doesn’t
14-May-09 www.steveransome.com Page 12
13. Comparing different module technologies
how important are any differences ? Diffuse Clear
Crystalline Silicon #1 and #3 Thin Film #4 and #6
Imax
(1)
(1) Imax quite similar, TF slightly worse current variability
– spectral mismatch/annealing
Vmax
(2)
(2) Vmax very similar, TF slightly better voltage thermal coefficients
14-May-09 www.steveransome.com Page 13
15. Insolation (kWh/m²/y)
vs Irradiance, Clearness index and Beam fraction
More Insolation at :
1. Higher irradiance
than lower –
most sites
2. Higher clearness
index (clear skies)
3. Higher beam
fraction (low
diffuse)
Irradiance (kW/m²) than at lower values
14-May-09 www.steveransome.com Page 15
16. Insolation (kWh/m²/y)
vs. Tmodule and Irradiance
Irradiance (kW/m²)
More Irradiance at high light levels
than low light even in Germany
Tmodule (C)
More frequent measurements show
even more high light level
17. How all weather parameters are correlated
making understanding data more complicated
Indoor (STC) Outdoor
<Worse weather Better weather>
Irradiance 1 kW/m² Lower Higher
Module temperature 25 C Colder Warmer
Spectrum AM 1.5 G Redder Bluer
Angle of incidence 0° normal Away from normal Nearer normal
Direct : Diffuse All Direct Mostly diffuse Mostly direct
14-May-09 www.steveransome.com Page 17
18. Extracting temperature coefficients
from outdoor data Imax and Vmax
Values may differ from internal
measurements as weather
parameters are correlated (e.g.
spectrum and temperature)
which will affect multi junction
thin film more than c-Si.
1. Vmax more accurate than
2. Imax
Need to filter out low
irradiance/temperature data
as too variable
14-May-09 www.steveransome.com Page 18
19. Empirical modelling
predicting Tmodule, Vmax and dc Power
• Simple empirical models
can predict Tmodule, Vmax
and Pmax
Tmod = f(Irrad, Tamb, WS, …)
Vmax = f(Irrad, Tamb, WS, …)
Pmax = f(Irrad, Tamb, WS, …)
• Can characterise measured
and predict future PV
performance
fits (black dots)
measured (coloured dots)
20. Empirical modelling
Flow chart
learning mode to
derive coefficients
Empirical formulae and
Inputs coefficients
Irradiance (kW/m²)
Ambient Temperature (C)
Cell Temperature C
Windspeed (ms-1)
MPP Voltage V
MPP Current A
MPP Power W
Report discrepancies Validate measurements
Sum (Power) = Energy Yield
22. Simulating outdoor performance,
extracting indoor parameters
indoors outdoors
Meas. DC module DC module String AC array
Stage (IV scan) (IV scan) (Vmp track) (Inverter)
Efficiency Pactual/ Module Inverter
vs nameplate, Mismatch efficiency,
Irradiance Dirt, Partial shading,
Temperature Thermal Wiring loss,
Spectrum Annealing, String mismatch
AOI etc. Degradation
Weather
Correlation
Parameter extraction Performance modelling
14-May-09 www.steveransome.com Page 22
23. Finding shading - Max. irradiance
per hour of the day and month of the year
• Good unshaded sites will
give smooth, symmetrical
oval shapes as shown
• Shading will show as lower
maximum irradiance than
expected for certain times
and months (e.g. after
14:00 November to January
for low horizon in the west)
14-May-09 www.steveransome.com Page 23
24. AC Performance : Maximum ac yield
per hour of the day and month of the year
• Well performing arrays will
give smooth, symmetrical
oval shapes as shown
• Thermal problems would be
seen by summer afternoon
dips (although this array
seems good)
• Turn on problems would be
seen by low values in the
morning
14-May-09 www.steveransome.com Page 24
25. Finding shading : Solar Irradiance and
array sum energy vs Solar position
Total insolation vs Solar
height and azimuth in 10° bins
Good unshaded sites should
have a symmetrical shape like
this in Germany
Horizon shading appears as
wide low irradiance areas
Tree or pole shading is seen in
tall low irradiance areas
14-May-09 www.steveransome.com Page 25
26. Finding “Turn on” problems and Shading
Performance/predicted Vs. Date and Time
(1)
1. Shading would show as
poor performance in
horizontal shapes
2. “Turn on” problems appear
as missing data at
beginning of day
(3) (2)
3. Missing data all day
27. Conclusions
• Sophisticated outdoor testing has been used to compare dc
modules with ac arrays
• Sum kWh/kWp figures alone are not enough to qualify
measurements
• A detailed knowledge of dc performance helps understand AC
data
• Normalise data for easier error checking V, I etc.
• Max. Irradiance or Power vs. time of day and month can
identify shading or thermal problems
• Checking raw data enables faults, limits and weather effects
to be analysed.
14-May-09 www.steveransome.com Page 27
28. Thank you for your attention
Thanks to ISET for the DC data
This paper and previous ones are available at
steve@steveransome.com
www.steveransome.com