Cultivation of KODO MILLET . made by Ghanshyam pptx
Core Objective 1: Highlights from the Central Data Resource
1. Core Objective 1: Highlights from
the Central Data Resource
Anubhav Jain w/
Robert White, Todd Karin, Mike
Woodhouse, & Cliff Hansen
DuraMat Fall Workshop, Sept 21 2020
2. The Central Data Resource develops and disseminates
solar-related data, tools, and software
DuraMat projects
generate data sets
Data enters the DataHub
and is access-controlled
by project
Users access data sets,
visual analysis tools, and
open-source software
3. Central Data Resource (Data Hub) objectives
DuraMat
Data Hub
Data Software
Analysis
Tools
• Objective: Collect and disseminate module
reliability related data, and apply data science to
derive new insights from data
• Key results include:
– A central data resource that securely hosts a
mix of private and public data of multiple data
types
– Development of open-source software libraries
that apply data analytics to solve module
reliability challenges
– Demonstrated use case of above tools to an
industrial use case
4. “A central data resource that securely
hosts a mix of private and public data of
multiple data types”
5. The Data Hub at a glance
137 Users
63 Projects
128 Datasets
70 Public Datasets
2120 Files and Resources
Google analytics July 1 –Aug 10, 2020:
https://datahub.duramat.org
6. How data is organized
Project Dataset Files or Resources
CSV
PDF
TEXT
XML
JSON
JPG
GIF
PNG
Excel
Team Member Access:
All project information and contained
datasets and files
Public Access:
Project names, descriptions and
abstracts
+ Links to
external data
63 available 128 available (70 public) 2120 available
Contact: Robert White
Data is PRIVATE by
default
Data can be
uploaded by
DuraMat project
members and
vetted sources
Data can be
accessed by
registering on
the Data Hub
web site
Data can be made public
through administrative
control, with
authorization
7. Highlights from current data sets
1. Albedo Data for Bifacial Systems
2. Coatings to Reduce Soiling and PID Losses
3. NREL Soiling map and supporting data
4. Bifacial Experimental Single-Axis Tracking Field data
For the public:
1. Back side defect imaging in crystalline silicon PV modules
2. Identifying Degradation Mechanisms in Fielded Modules using
Luminescence and Thermal Imaging
3. Combined Accelerated Stress Testing
4. Effect of Cell Cracks on Module Power Loss and Degradation
For project members
8. Highlights from current data sets
1. Albedo Data for Bifacial Systems
2. Coatings to Reduce Soiling and PID Losses
3. NREL Soiling map and supporting data
4. Bifacial Experimental Single-Axis Tracking Field data
For the public:
1. Back side defect imaging in crystalline silicon PV modules
2. Identifying Degradation Mechanisms in Fielded Modules using
Luminescence and Thermal Imaging
3. Combined Accelerated Stress Testing
4. Effect of Cell Cracks on Module Power Loss and Degradation
For project members
See also:
Poster C01 -2
from Robert White
9. “Development of open-source software
libraries that apply data analytics to solve
module reliability challenges”
10. Technology Summary and Impact
Resources
Automatic Crack Detection using Convolutional Neural Networks
• Take in electroluminescence images of
full modules, automatically crop out cells,
and identify cracks and power loss
regions
• Working with EPRI to correlate cracks
with power loss
• Testing on diverse images with PVEL
• https://github.com/hackingmaterials/pv-vision
• Poster presentation by Cara Libby in CO3-3:
“Effect of Cell Cracks on Module Power Loss &
Degradation”
Cracks, defects, and other features predicted by U-Net machine
learning model
Busbar detection (gold)
Cracks (purple)
Power loss regions
(green)Cells in module are
automatically detected,
cropped out, and
perspective corrected
Contact: Xin Chen
11. Technology Summary and Impact
Resources
Detecting changes in module parameters using production data sets
• Goal: Use operating / production data
(e.g., Vmp and Imp, and Tcell) to determine
changes in module parameters (e.g.,
Rseries and Rshunt) over time
• Method based on “Suns-Vmp”1
• Compare method with systems for which
detailed diagnostics are available, e.g.,
NREL SERF East
• https://github.com/DuraMat/pvpro [[in development]]
• Poster presentation by Todd Karin in CO1-4:
“Introduction to PVPRO”
• Next talk in this workshop
By analyzing changed in Vmp and Imp throughout the course of
deployment, PV-PRO aims to detect changes in module parameters
Contact: Todd Karin
arrays. But about a fifth of the observed changes were from
the inverter not tracking the peak-power as effectively as the
PV arrays aged.
1. Background
As part of the construction of the Solar Energy Research
Facility building at the National Renewable Energy
Laboratory in Golden, Colorado, two grid-connected
photovoltaic systems were installed on the roof to provide
power to the building and the utility grid. Corresponding to
their location on the building, the systems are identified as
SERFEAST and SERFWEST. The SERFEAST PV array is
shown in Fig. 1.
Figure 1. SERFEAST array on the roof of the building.
Each PV array consists of 140 Siemens Solar Industries
model M55 PV modules. The PV arrays are electrically
connected as five source-circuits, with each source-circuit
having a positive and negative monopole of 14 series-
connected PV modules. Each PV array is connected to an 8
kW Omnion Series 2200 inverter for conversion from d.c. to
a.c. power. The PV arrays are tilted from the horizontal at
an angle of 45 and are aligned with the azimuth of the
building that is oriented 22 east of south. The longitude and
degradation over the 8-year period.
2. Data Screening
For calculating PV system ratings, data were selected to
meet meteorological criteria and to avoid data recorded
when the inverters were malfunctioning or off-line for
repairs.
Meteorological criteria for data selection were a 15-
minute average POA irradiance greater than 800 W/m2 and
an angle-of-incidence of direct-beam radiation to the PV
array of less than 30 degrees. This ensures that the cloud
presence was small and that the pyranometer measurement
of irradiance was performed within a range of incident
angles where the cosine response of the pyranometer is not
detrimental to measurement accuracy.
A region of acceptable PV array operating voltages as a
function of PV array temperature was identified using data
recorded during normal system operation. This resulted in
the “boxed” area shown in Fig. 2. Data within the “boxed”
area were judged acceptable for use for data analysis,
whereas data in the remaining area were judged
unacceptable because they were measured under
malfunctioning or system-off (open-circuit) conditions.
Normal operation for these systems does not necessarily
mean peak-power tracking, although that was the original
intent. The inverters were specially ordered to achieve a
peak-power tracking range of 200 to 280 volts. However, as
delivered, the inverters do not operate below about 220
volts. Consequently, for elevated PV array temperatures, the
inverters do not peak-power track because the PV arrays are
operated at 220 volts and the PV array voltage for maximum
power (Vmp) is considerably less.
The diagonal lines in Fig. 2 represent PV array Vmp values
as a function of PV array temperature for 1994 and 2002.
They were determined from PV module and array current-
voltage (I-V) curve measurements. Values of Vmp for 2002
are about 10 volts less than they were in 1994;
consequently, for elevated PV array temperatures in 2002,
the inverter operates the PV array further from its peak-
power point than in 1994. As an example, the power penalty
for not peak-power tracking at a PV array temperature of
1
Fig. 6. Pm ax , Isc , Voc , and FF degradation for all measured strings and
arrays. The East array is listed on top and the West array at the bottom. The
string polarity is indicated by negative (N) and positive (P).
and not voltage. The fairly large uncertainties are caused by the
multiple data shifts that needed to be corrected.
IV. OUTDOOR I–V MEASUREMENTS
A total of eight sets of I–V measurements were taken dur-
ing the 20-year lifetime of the system. The measured module
temperatures were translated to 45 °C, which presented a good
approximation to the average temperature for this particular lo-
cation, and irradiance to 1000 W/m2
. It was not clear whether a
linear or nonlinear regression resulted in a better fit to the data.
Thus, in the absence of a clear indication, a linear regression line
was used through the eight data points for each string, subarray,
and array [8], [9]. The resulting degradation for each parameter,
subarray, and string are summarized in Fig. 6. Maximum power
(Pmax) is indicated by blue circles, short-circuit current (Isc)
by red squares, open-circuit voltage (Voc) by green triangles,
and fill factor (FF) by inverted purple triangles. The strings for
the East array are shown on top and those for the West array at
the bottom. The uncertainty bars are statistical uncertainties cal-
culated from the regression standard errors. For the East array
(top), the Pmax degradation for the strings of negative polarity is
between 0.4%/year and 0.6%/year with the exception of string 3.
String 3 shows a higher degradation rate of about 0.8%/year that
seems to determine the overall degradation of the negative sub-
array. The decline appears to be dominated by FF decline for
this particular string, which is typically associated with reduced
shunt resistance or increased series resistance [10]. Increased
series resistance for aged PV systems is often caused by flawed
solder interconnects in combination with thermal cycling and
manifests itself by localized hot spots [11]. Less of the decline
can be attributed to Isc degradation, which is typically associ-
ated with light-induced degradation, discoloration, delamina-
tion, and soiling [12], [13].
Fig. 7 shows optical and infrared (IR) images of observed
discoloration, soiling, and local hotspots, visually corroborating
the I–V analysis. Nevertheless, the discoloration appears to be
less than in hotter climates for similar modules [14]. The overall
Fig. 7. Optical images of the system show some discoloration in the center
of most cells (a), permanent soiling (b), and some hotspots in IR imaging (c).
Photos (a), (b), and (c) by D. Jordan, NREL.
subarray degradation of about 0.7%/year is slightly less than the
average published literature Pmax degradation of 0.8%/year [3].
Historical degradation is more dominated by the Isc decline of
about 0.5%/year and 0.3%/year of FF, while the decline for this
particular system is more mixed or even more FF attributable.
Similarly to historical degradation, Voc degrades the least.
For the positive polarity, the Pmax degradation is more evenly
spread between the individual strings leading to a degradation
of the subarray of about 0.7%/year. Strings 3–5 are dominated
by FF losses, while strings 1 and 2 are characterized by more
dominant Isc losses. For the West array with negative polar-
ity, most strings degrade in Pmax in the 0.5–0.6%/year range.
Only string 2 degrades in the 0.7%/year range, thus apparently
determining the overall degradation for the subarray. Strings 2
and 3 show an equal degradation that is attributable to Isc and
FF decline. The other strings show more dominating FF losses.
The positive polarity of the West array shows the overall highest
Pmax degradation, which seems to be significantly influenced
by string 5. That is also the string that shows significant Voc
losses compared with all other strings.
Hotspots / Rs increase
over time
1. Sun, X., Chavali, R. V. K. & Alam, M. A. Prog Photovolt Res Appl 27, 55–66 (2019).
12. Technology Summary and Impact
Resources
A Quick and Easy to Use LCOE calculator
• Provide a visual, user-friendly tool for
quick “back-of-the-envelope” of LCOE
• Make rough estimates such as “if I deploy
a coating that increases cost by X, how
much additional efficiency do I need to
justify the cost?”
• Many preset options that are easily
tunable / configurable
• https://github.com/NREL/PVLCOE
• https://www.nrel.gov/pv/lcoe-calculator/
• Poster presentation by Brittany Smith in CO1-3:
“Presentation of Fiscal Year 2020 Results from
Technoeconomic Analysis for DuraMAT”
NREL’s online LCOE calculator allows users to quickly compare the
LCOE of proposed systems against a baseline
Contact: Brittany Smith
13. • Software and algorithms for data cleaning
– https://github.com/pvlib/pvanalytics [[POSTER: CO1-1, Hansen]]
• PV-Terms - a project to unify terminology in software
– https://github.com/DuraMAT/pv-terms
• Integrating PVDAQ data sets into DataHub
– https://pvdata.duramat.org
• Tools for string length calculations
– https://pvtools.lbl.gov/string-length-calculator
• Climate descriptors potentially relevant to solar degradation
– https://pvtools.lbl.gov/pv-climate-stressors
• Specific techno-economic studies conducted with industry partners
– [[POSTER CO1-3, Woodhouse & Smith]]
Other Central Data Resource Projects (historical and current)
14. • The Central Data Resource aims to
maximize the potential of applying
data as a resource to help solve
problems related to solar degradation
• We would be happy to hear any ideas
you might have for how to extend this
initiative!
Conclusion
DuraMat
Data Hub
Data Software
Analysis
Tools