More Related Content
Similar to Determining the causes and rates of PV degradation using the Loss Factors Model (LFM) with high quality IV measurements (20)
More from Sandia National Laboratories: Energy & Climate: Renewables (20)
Determining the causes and rates of PV degradation using the Loss Factors Model (LFM) with high quality IV measurements
- 1. www.steveransome.com25-Oct-16 1© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Determining the causes and rates of PV degradation
using the Loss Factors Model (LFM)
with high quality IV measurements
Steve Ransome1 & Juergen Sutterlueti2
1Steve Ransome Consulting Limited, London UK
2Gantner Instruments, Germany
PVPMC #6 – Freiburg Germany
25th Oct 2016
- 2. www.steveransome.com25-Oct-16 2© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Introduction to degradation analysis
• Most reported PV degradation are STC corrected efficiency only
(1kW/m2, 25C, AM1.5, AOI=0, direct only)
• ISC variability dominates performance uncertainty
(due to soiling, spectral effects, irradiance sensor calibration …)
Can we analyse other parameters independently of ISC ?
• The cause of degradation (e.g. RSHUNT, RSERIES, VOC )
gives site dependent energy yield degradation rates
(due to differing proportions of insolation vs. irradiance, TMOD etc.)
- 3. www.steveransome.com25-Oct-16 3© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Smooth IV curves are needed for good RSC and ROC calculations
“Rn = Apparent resistance between adjacent data points”
Typical GI measured IV curve (CdTe) GI raw measured (smooth data) vs.
synthesised “poor” data
truncated accuracy and added noise
𝑹 𝒏 = −
∆𝑽
∆𝑰
= −
𝑽 𝒏 − 𝑽 𝒏−𝟏
𝑰 𝒏 − 𝑰 𝒏−𝟏
Worse RSC accuracy from synthesised data
(e.g. truncated or noisy)
ISC, RSC
VOC, ROC
- 4. www.steveransome.com25-Oct-16 4© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Checking IV data quality with Log Resistance-Voltage (RV) curves
GI data much smoother than NREL’s Daystar and therefore easier to fit.
Can ignore a few “bad end points” with V~0 or V > VOC
GI
CdTe
NREL
CdTe
- 5. www.steveransome.com25-Oct-16 5© SRCL / Gantner Instruments
PVPMC #6 Freiburg
SRCL/Gantner “Loss Factors Model” [LFM]
GI data
Measure raw IV curves = f(G,T)
Fit lines to RSC and ROC
Normalise data to datasheet
6 normalised losses LFM
PRDC = nISC*nRSC*nIMP * nVMP*nROC*nVOC
Cell mismatch,
shading
Cell rollover
Curvature for better understanding
- 6. www.steveransome.com25-Oct-16 6© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Comparing “Loss Factors Model” with standard models
L
F
M
1-diode
(and similar
models)
Fit to IV curves Exact values for 8 parameters
around every IV curve
“Best fit to whole curve” depends on
data point distribution/weighting
“imperfect traces” e.g. cell mismatch, roll over
Normalised values for
module variability
Yes.
e.g. “nRsc = 98.0 ± 2.0%”
No. Specific module data only
e.g. “RSHUNT = 1234Ohms”.
Independent
Parameters ?
Almost independent
(nVOC depends a little on nRSC and nISC)
No. Parameters are often interdependent e.g.
nF and Io
Dependence on
Irradiance and
temperature
Simple optimum fits give
exact coefficient behaviour for low light,
temp coeffs etc. each module
Try to fit pre-defined equations (even if they
don’t fit data) e.g. RSHUNT (GI), I0(TCELL) etc. low
light and temp coeffs. may be wrong.
Separation of all inputs
e.g. ISC ~ AOI, SR
Not needed. Can just measure outdoor params
(for ISC separate clear from cloudy skies)
Need to separate all parameters
ISC = ISC0 * f(AOI) * f(SR) …
Fault finding and
quantification of loss
Yes. Can easily identify quantify
Cell mismatch, shading, R and VOC changes etc.
Some are possible (e.g. RSHUNT, RSERIES)
but not mismatch, rollover etc.
- 7. www.steveransome.com25-Oct-16 7© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Yearly
IV traces
by
irradiance
Sept.
2010-16
GI data
Discrepancies
seen at very low
light levels ?
Changes in LFM parameters
ΔnISC ΔnRSC
ΔPRDC
ΔnROC
ΔnVOC
For each module and Irradiance (e.g. ~0.8kW/m²)
ISC variability e.g. soiling, sensor calibration etc.
- 8. www.steveransome.com25-Oct-16 8© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Yearly
IV traces
by
irradiance
Sept.
2010-16
GI data
Discrepancies in
ISC seen at very
low light levels
0.04kW/m²
Why ?
Year Deg
%/y
If module is
degrading it’s
worse at low
light
1.0 0.3
kW/m2
- 9. www.steveransome.com25-Oct-16 9© SRCL / Gantner Instruments
PVPMC #6 Freiburg
GI Tempe OTF from North to South East
Low horizon shading for morning sun
GI hut position red
Power lines green
Sensors Cyan
Modules Magenta
Google Street view from south east
- 10. www.steveransome.com25-Oct-16 10© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Shading from powerlines affect the sensors and modules at
different times of morning (5 distinct dips)
Efficiency Isc / Gi
Approx. shade times
modules (07:35-08:05)
sensors (07:10-07:40)
Sensors higher than
modules so are shaded
earlier in morning
(Late afternoons are
affected by 2D tracker)
Low light performance
measurements vs.
irradiance must be
properly corrected for
shading
- 11. www.steveransome.com25-Oct-16 11© SRCL / Gantner Instruments
PVPMC #6 Freiburg
SRCL/Gantner “Loss Factors Model” vs. Irradiance
detailed information at www.steveransome.com, GI data
•A drop in any LFM
parameter limits
overall PRDC
•Any LFM parameter
changing over time
affects PRDC
PRDC = nISC*nRSC*nIMP * nVMP*nROC*nVOC
Low light
limiting
High light
limiting
- 12. www.steveransome.com25-Oct-16 12© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Analysis method for frequent IV curves outdoors
GI Data
Sudden change –
damaged or failed
module
Steady decline
module
Stable performance
module
PRDC from 6 years of hourly measurements 2010-2016
modules chosen
to analyse
differing behaviour
PRDC at Low light may be
seasonally dependent (longer day length,
sun behind module)
PRDC at High Irradiance tends not to
be seasonally dependent
- 13. www.steveransome.com25-Oct-16 13© SRCL / Gantner Instruments
PVPMC #6 Freiburg
LFM
vs.
irradiance
GI data
It’s hard to see any
changes in nISC
unless corrected
for shading,
soiling, aoi, sr and
direct:diffuse
- 14. www.steveransome.com25-Oct-16 14© SRCL / Gantner Instruments
PVPMC #6 Freiburg
LFM
vs.
irradiance
GI data
Irradiance
dependent
degradation
dnRSC
Irradiance
independent
degradation
dnROC
- 15. www.steveransome.com25-Oct-16 15© SRCL / Gantner Instruments
PVPMC #6 Freiburg
nRsc vs. DateTime and Log(Irradiance)
Low light levels performance degrades much faster than high light levels
High light levels
(0.5–1.0kW/m²)
dnRSC -0.5%/y
Low light levels
(0.1–0.2kW/m²)
dnRSC -2.0%/y
Very Low light levels
(0.001–0.02kW/m²)
dnRSC -5%/y
- 16. www.steveransome.com25-Oct-16 16© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Measurement Conclusions
NOTES:
• Atypical devices analysed vs. a stable module
• Smooth IV curves needed for degradation analysis
(check if “Rn = –V/I” is good on your measurement system)
GANTNER INSTRUMENTS dataset in AZ (6 years) - SRCL/GI Loss Factors Model
• LFM separates degradation components from nISC
• Good Gantner Instruments IV trace quality allows study of RSC and ROC
• Modules may degrade differently at high or low light levels
• LFM allows a fast independent check of degradation rates
- 17. www.steveransome.com25-Oct-16 17© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Predictions : Site Dependent Energy Yield Degradation
Energy Yield Gi,Tmod [Insolation(Gi,Tmod) * Efficiency(Gi,Tmod)]
Irradiance distribution is
site dependent
(cumulative Hi kWh/m² % > Gi kW/m²)
nRSC (related to RSHUNT)
degradation/y vs. Irradiance
-2.0%/year
low light
-0.5%/year
high light
*
- 18. www.steveransome.com25-Oct-16 18© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Predictions : Energy yield degradation rate
at sites (from measured dnRSC )
High Insolation site
=
Lower Energy Yield
degradation
-0.7%/y
Lower Insolation site
=
Higher Energy Yield
degradation
-1.3%/y
- 19. www.steveransome.com25-Oct-16 19© SRCL / Gantner Instruments
PVPMC #6 Freiburg
Predictions : Conclusions
LFM gives
• Degradation rates for various
parameters vs. irradiance etc.
• Predicted Energy Yield (kWh/y)
degradation vs. site
• Low light drops in nRsc (~ RSHUNT)
cause worse falls at low than high
insolation sites
• Analysis methodology is being
integrated into
• www.gantner-webportal.com
(see separate poster)
Thank you for
your attention!