This document discusses systematic approaches for ensuring the correct representation of measured multi-irradiance PV module performance in PV system energy production forecasting software. It presents methods for optimizing the default module models in PV*SOL and PVsyst to measured data from a set of Yingli Solar mc-Si PV modules. Optimizing the models reduced the RMSD between modeled and measured performance from 6.19% to 0.05% in PV*SOL and 3.46% to 0.07% in PVsyst. Energy forecasts using the optimized models showed gains of 3-6% over default models in PV*SOL and 2-3% in PVsyst across four locations. The document advocates characterizing modules per

1. Kenneth J. Sauer1, Thomas Roessler2
1Yingli Green Energy Americas, Inc.
2Yingli Green Energy Europe GmbH
Presented at the 2013 PV Performance Modeling Workshop
Santa Clara, CA | May 1-2, 2013
Systematic Approaches to Ensure
Correct Representation of Measured
Multi-Irradiance Module Performance
in PV System Energy Production
Forecasting Software Programs
Published by Sandia National Laboratories with the permission of the authors.

2. 2
2
Outline
• Basics
– Main components of PV module models
– Multi-irradiance behavior of PV modules
– Self-reference method
• Multi-irradiance behavior in PV*SOL® and PVsyst
• Motivation: Example data and default models
• Systematic approaches for fitting PV module models
‒ PV*SOL®
‒ PVsyst
• Impact on energy forecasts
• Summary and outlook
• Appendix: New software versions

3. 3
3
Basics: Main components of PV module models
Electrical parameters at
STC (I-V curve)
Multi-irradiance behavior
Temperature coefficients
(and NOCT)
Incidence Angle Modifier
(IAM)
Module
model
Provision via data sheet Additional characterization

4. 4
4
Basics: Multi-irradiance behavior of PV modules
• Characterized via “relative efficiency deviation” ∆rel [%]
as function of irradiance G [W/m2]:
Pmax(G) Gref
Pmax,ref G
Subscript “ref” means reference conditions; usually STC (where Gref=1,000W/m2).
• Typical
behavior:
∆rel (G) = x ‒ 1
-10%
-8%
-6%
-4%
-2%
0%
2%
0 200 400 600 800 1000
∆η
rel
G [W/m2]

5. 5
5
• For linear devices, if data are obtained from indoor flash testers,
check that deviations from Isc linearity are <1% (see IEC 61853-1
Section 8)
• Possible causes of Isc non-linearity include but are not limited to:
• Spectral mismatch between the irradiance sensor and sample under test if the
spectrum changes over the irradiance range
• Spatial non-uniformity caused by filter placement during calibration or test
• Obtain effective irradiance G
used to calculate Δηrel following
the self-reference method (IEC
61853-1 Section 8):
G = Gref * (Isc / Isc,ref).
Basics: Self-reference method
-5.0%
-4.0%
-3.0%
-2.0%
-1.0%
0.0%
1.0%
2.0%
200 400 600 800 1000
Δη
rel
G [W/m2]
Deviation from Isc linearity
No adjustments
Self-reference method

6. 6
6
PV*SOL® (v4.0)
• Multi-irradiance behavior via one value
of irradiance below reference conditions
(Gpartial_load), at which I-V curve
parameters must be provided
PVsyst (V5.56)
• Multi-irradiance behavior via 5 parameters
of one-diode model (series and shunt
resistance, photo current, saturation current
and diode quality factor), where 2 additional
parameters are used to describe RSh(G)
Multi-irradiance behavior in PV*SOL® and PVsyst
Gpartial_load
RSh, RS

7. 7
7
-35%
-30%
-25%
-20%
-15%
-10%
-5%
0%
5%
0 200 400 600 800 1000
∆η
rel
G [W/m2]
Default model behavior, PV*SOL®
Default model behavior, PVsyst
Measured averages
Motivation: Example data and default models
• (22) 235W Yingli Solar mc-Si PV modules
• Reputable 3rd party laboratory in USA
• Class AAA pulsed solar simulator
• I-V curve at STC
• 4 additional I-V curves (25°C and AM 1.5g)
– 800W/m2
– 600W/m2
– 400W/m2
– 200W/m2
• Δηrel(200W/m2):
• Measured averages = -2.9%
• Default PVsyst (V5.56) = -8.3%
• Default PV*SOL® (v4.0) = -14.5%
IEC 61853-1
Section 8.5

8. 8
8
Systematic approach for
PV*SOL®
• One user-modifiable variable
(Gpartial_load), thus manual
optimization possible
• Limitations of coarse-grained
grid of tested G
• Interpolation of measured
data (IEC 61853-1 Section 9)
• RMSD: 6.19% (default) 
0.05% (optimized)
76.90%
76.95%
77.00%
77.05%
77.10%
0.00%
0.05%
0.10%
0.15%
0.20%
0.25%
0.30%
0.35%
0.40%
280 290 300 310 320 330 340 350
FF
RMSD
G [W/m2]
RMSD, PV*SOL® Optimization
FF, Interpolated Measured Averages
-35%
-30%
-25%
-20%
-15%
-10%
-5%
0%
5%
0 200 400 600 800 1000
∆η
rel
G [W/m2]
Default model behavior,
PV*SOL®
Optimized model behavior,
PV*SOL® (400W/m²)
Optimized model behavior,
PV*SOL® (320W/m²)
Measured averages

9. 9
9
Systematic approach for
PVsyst
• Optimization over a 4-
dimensional space of
parameters describing
parasitic series and shunt
resistances
‒ Numerical optimization
procedures are required
• RMSD: 3.46% (default) 
0.07% (optimized)
-35%
-30%
-25%
-20%
-15%
-10%
-5%
0%
5%
0 200 400 600 800 1000
∆η
rel
G [W/m2]
Default model behavior, PVsyst
Optimized model behavior, PVsyst
Measured averages

10. 10
10
• Annual energy
production simulated
over four geographically
distributed locations
• Reported as gains in
forecasted outputs
using optimized models
over forecasted outputs
using default models
– PV*SOL®: 3-6% gains
– PVsyst: 2-3% gains
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
6.0%
Toronto Austin Berlin Rome
Gain
in
Forecasted
Annual
Energy
Output
PVsyst
PV*SOL®
Impact on energy forecasts

11. 11
11
Summary and outlook
• Risk of relying on default settings that prescribe irradiance
behavior in PV*SOL® and PVsyst module models
• Importance of characterizing PV modules following, e.g.,
IEC 61853-1
– Plus fine-grained measurements close to max[FF(G)]
• Systematic optimization procedures reported for fitting PV
module models to measured data
• Next step: field validation of optimized models

12. 12
12
-35%
-30%
-25%
-20%
-15%
-10%
-5%
0%
5%
0 200 400 600 800 1000
∆η
rel
G [W/m2]
Default model behavior, PVsyst V5
Default model behavior, PVsyst V6
Optimized model behavior, PVsyst
Measured averages
-35%
-30%
-25%
-20%
-15%
-10%
-5%
0%
5%
0 200 400 600 800 1000
∆η
rel
G [W/m2]
Default model behavior, PV*SOL® v4.0
Default model behavior, PV*SOL® v5.5
Optimized model behavior, PV*SOL® (320W/m²)
Measured averages
Appendix: New software versions
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
6.0%
Toronto Austin Berlin Rome
Gain
in
Forecasted
Annual
Energy
Output
PVsyst V5
PVsyst V6
PV*SOL® v4.0
PV*SOL® v5.5