21 gi 8th_pvpmc_talk_abq_final_ext_170508t19_new

www.gantner-environment.comPage 1
Confidential|SubjecttoNDA
Page 1
Optimized PV performance using state of the art monitoring
for increased asset value
8th PV Performance Modelling and Monitoring Workshop
Albuquerque; May 10th, 2017
5/18/2017Juergen Sutterlueti et al.

Outline
1
5/18/2017
Market trends
2 Gantner PV Business
3 Building blocks for “PV asset monitoring”
4 Real time data processing
5 Prediction (Mechanistic Power Model)
6 Plant Control – supportive functions for the grid
7 Outlook & Next Improvements

Source IRENA, 2016,
Price erosion of Cost per MWh of PV Electricity
Tender PV Power Purchase Agreements (PPAs) between 2010 and 2016
How to measure PV
performance and mitigate risk?
5/18/2017
PV Market:
• Globally distributed
• PV market conditions are harsh &
dynamic
• Extreme cost pressure vs.
“similar quality”
But still:
• PV Assets are an attractive
strategic investment and stable ROI
is expected
• PV Monitoring is the main source
for Quality status and Risk
reduction tool.

Gantner PV Business
5/18/2017
• A world leader in independent PV performance
monitoring equipment and services
• 3.6 GW sold, and over 110 customer projects being
monitored by Gantner
• Recent new customer contracts in key emerging
markets, e.g. Egypt, Jordan
• Global R&D partnerships with Sandia National
Labs, Fraunhofer, TUV Rheinland, et al.
• Over 32 local reps and 10 offices for Level 1 support
worldwide;
• Outdoor test facility for PV module characterization
and analysis, located in Arizona, US

gantner.webportal – Software as a Service Solutions
Main benefits
• One unified platform for effective and automated PV
plant / Asset management
• Vendor-independent, scalable PV plant monitoring
• Real time data handling with advanced analytics and
asset benchmark, KPIs
• Effective O&M handling (Alarms, Document handling,
issue management, reporting) reduces risk and O&M
costs; IEC 61724 Ed.2:2017 ready
• Optimized PV production leads to increased asset value
and reduced risk
• Provides baseline for energy prediction (day ahead),
energy trading, etc.
• Independent Data Archive Reduced Risk of Stranded
Data, Increases Project Bankability
5/18/2017

Building blocks for “PV asset monitoring”
(Analog) big data in context of Photovoltaics
• Make it traceable
• Do harmonized processing
• Make it structured – but still agile
• Allow real time operations
(aggregation, filters, benchmark, corrections)
• Comparison: Target vs. actual, prediction
5/18/2017

Data processing – real time
E.g: 5000 channels per 10 MW to import and process
5/18/2017
Fast check to validate all
measurements sensors,
inverters, strings etc. – first
as sensible, then good.
Automated checks
(real time), constant
performance
Regular
sanity
checks
Multiple import
data streams,
standardized

Traceable data management and import
Data processing - real time
REAL time platform
5/18/2017
Multiple sources need harmonized data concept
• Naming convention:
▪ For fast and easy use
▪ Structure: {Parameter}_{Modifier}-{Component}-{ID}
Example:
Pdc -Inv-1.1.1-1,
Pdc_T-Inv-1.1.1-1
Gi-GiPyr-1.1
Idc-Mpp-1.01.01.03.1
• Color code & Unit convention
• Filters/Limits:
▪ Run all data sets with
standardized filters
• Converter for 3rd party

Structure view
5/18/2017
• Automatic (visual) hint where issues are
happening
▪ Bold lines show problems
▪ Warning: Orange, Alarm: Red,
No issues: Green
▪ Fast validation check of all
measurements (sensors, inverters,
strings)
• Possible by Inventory concept and Naming
convention
• Allows configuration changes
• Good feedback during plant startup and
commissioning (limits check, .. )
• Automatically rendered Structure
• Real time calc from Module to Site Level

Real time data processing
- Live Demonstration -
(see Appendix for examples)
5/18/2017

Real time calculation: Aggregation
Aggregation - from PV Module up to site level
5/18/2017

Real time calculation: Normalization
Normalization / Deviation of Strings, SCB, Inverter
5/18/2017
Absolute Values:
I MEASURED differs for larger and smaller arrays
Normalized values:
IMEASURED/INOMINAL show similar values

Real time calculation: Benchmark I
Different assets
5/18/2017

Real time calculation: Benchmark II
Different assets
5/18/2017

Real time calculation: Benchmark – Table view
5/18/2017

Real time calculation: Prediction I
Mechanistic Power Model (all levels): Target vs. Actual
5/18/2017
more details :
Ransome, S.; Sutterlueti, J.:
„Choosing the best Empirical Model for predicting energy yield”; 7th
PVPMC Workshop, SUPSI, Canobbio Switzerland, March 2017
https://www.slideshare.net/sandiaecis/15-2017-pvpmc7ransome170330t081corrected2-
74980695
We invite all interested people to join us in developing this further and also
add it to IEC 61853, we are already working with several 3rd parties.

PMEASURED /
PNOMINAL

PFIT /
PNOMINAL
Difference

Real time calculation: Prediction II
Mechanistic Power Model (e.g. at all levels): Target vs. Actual; different levels
5/18/2017

CONTROL level
Plant Control – supportive functions for the grid
Example Reference site in UK and R&D project
• Control Function and date interface is a must for
any inverter/storage concept
• Open interfaces and protocols are key to enable
efficient Operation
• DNO requirements can all be fulfilled
• Project:
▪ DNO: UK Power Networks, UK
▪ Size: 45,6 MW AC
▪ 1122 Inverters, 4 grid connection points
• Full monitoring and control
▪ Absolute production constraint
▪ Power gradient constrain
▪ Voltage Control
▪ Reactive Power Control
▪ Power Factor Control
▪ Frequency Control
▪ Ramp rate limitation
• Gantner Control is part of:
▪ ReWP – „Regelleistung durch Wind- und
Photovoltaikparks“ (72.6MW, 15 plants)
▪ Members:

Outlook & Next Improvements
• Improved operations and maintenance (O&M) capabilities
• Estimation of the degradation rate
• Detection and classification of failures, result of “Innovative Performance Monitoring System for Improved
Reliability and Optimized Levelized Cost of Electricity” (IPERMON), with University of Cyprus
• Capacity testing
• PV production forecasting (component level)
Software as a Service
5/18/2017

Outlook: Ready for Full Digitalization?
• for future technology applications
▪ IoT, IoE, AI, Big Data analysis, Neural networks, preventive O&M
▪ PlugIn ready
• Cloud (public, private, local cloud), scalable
• Open interfaces:
▪ can all other devices: Ready for the future
• 3rd party hardware I/O integration
• New and
• Options for paid services as
▪ Pushes applications, data and computing power (services)
to the logical extremes of a
network.
▪ Edge computing layer
compute, storage, and networking.
▪ Units will be which
becomes the primary job of the device.
Device and platform requirements for Digitalization

Summary
Asset Monitoring Solutions should consist of
• Independent & open interfaces
• Granular data collection
• Harmonized & traceable data sets (secondary market sales)
• Easy to use data analysis tools, including financial KPI’s
• Provide tools for O&M Engineers and Investors/Owners
• One unified platform for effective and automated PV plant management
• Investor's profit is dominated by OPEX
 O&M strategy and efficiency is key success factor (Post-FiT/ITC market)
• Metrics will focus on “How much was Not produced" when
you focus on IPP, PPA or NetMetering projects
5/18/2017

Thank you very much for your attention!
Sample projects out of 3.6 GWp with turn key Gantner Instruments HW & SaaS Solutions incl. Grid Control
We would like to thank the Gantner team for their indefatigable effort in realizing successful customer projects globally.
www.gantner-webportal.com j.sutterlueti@gantner-instruments.com
5/18/2017

Appendix
5/18/2017

About Gantners PV Business Unit
Gantner’s PV Business Unit has quickly grown to become one of the leading
independent providers of performance monitoring equipment and services for
utility-scale PV projects in many markets around the world.
The company currently boasts over 110 projects (3.6GW) of PV projects installed
with Gantner monitoring equipment.
PV customers include developers, EPC’s and investors, and all have come to
appreciate that Gantner’s equipment enables superior performance monitoring that
results in significant financial benefits for the projects, including lower O&M costs,
lower LCoE and higher ROI.
5/18/2017

Products and Services
Robust Hardware and Software for Optimized
Control
• Based on proven Gantner architecture and
decades of experience in many industries
• Wide range of products to provide customer
flexibility: DG or Central, combiner boxes, data
logger, controller
• SCADA, String, Combiner Box or Project Level
• Independent & open interfaces
• Independent web-portal solution
• Granular data collection
• User-friendly data analysis tools, including
financial KPI’s
Robust Hardware and Software for Optimized Control
5/18/2017

String Level Products
5/18/2017
• Identify design and production
errors quickly – inverter
malfunction, soiling, shading, PV
module degradation
• DC shunts provide 10x more
accurate current measurements
• Up to 1500V
• Consume < 1.5W
• Inverter independent
• More than 2GW installed
String.bloxx Series
DC and AC Combiner Boxes

Data Level Performance Monitoring
5/18/2017
• One device can handle up to 100MWdc
• Signal conditioning, data storage,
communication (Analog inputs, Digital I/Os, RS-485
Ports, Ethernet interface, Industrial data memory, Touch
screen display)
• Collect all string-level data, plus met data
and key equipment status
• Support protocols from all leading
inverters
• Comply with grid interface standards (IEC
61724 Ed 2, DNP3)
• Ready for NEW IEC 61724-1:2017

CONTROL level
Plant Control – Same for Central or Decentral
Example Reference site in UK and R&D project
• Control Function and date interface is a must for
any inverter/storage concept
• Open interfaces and protocols are key to enable
efficient Operation
• DNO requirements can all be fulfilled
• Project:
▪ DNO: UK Power Networks, UK
▪ Size: 45,6 MW AC
▪ 1122 Inverters, 4 grid connection points
• Full monitoring and control
▪ Absolute production constraint
▪ Power gradient constrain
▪ Voltage Control
▪ Reactive Power Control
▪ Power Factor Control
▪ Frequency Control
▪ Ramp rate limitation
• Gantner Control is part of:
▪ ReWP – „Regelleistung durch Wind- und
Photovoltaikparks“ (72.6MW, 15 plants)
▪ Members:
PlantControl

gantner.webportal – Software as a Service Solutions
Main benefits
• One unified platform for effective and
automated
PV plant management
• Vendor-independent PV plant monitoring
• Fully compliant w/ international standards
like IEC 61724 Ed.2
• Comprehensive financial KPIs for Investors
• Optimized PV production leads to increased
asset value and reduced risk
• Provides baseline for energy prediction (day
ahead), energy trading, etc.
5/18/2017
SCADAlevel

Summary
• Data Requirements Are Growing – PV Market Expansion, PV Grid Integration
• Reliable Data Drives Value
▪ DC shunt architecture provides 10x more accuracy
▪ Lower O&M, Increased Production, Lower LCoE, Higher ROI
• Gantner’s Open Interface Provides Versatility to Project Owners, Operators and Grid
Operators
• Q.Reader provides single platform for grid integration – both monitoring and external PV
power plant control
• Independent Data Archive Reduced Risk of Stranded Data, Increases Project Bankability
• Strong Track Record – 3.6GW being monitored worldwide, best-in-class customers and
partners
5/18/2017
Strategy

Technical Features (selection) I
Functions (selection)
5/18/2017
• Data consistency
• Perform real-time sanity checks for physical meaningful
values
• Uptime tracking / Availability audit
• Provide overall Sensor review for measurement validity
• Monitor Irradiation sensor drift
• All incoming data is tested for measurement-, data
consistency or data transfer issues
Filters
• New Filter Parameters: Sun position, AOI, Shading
status, Clearness Index (kTh), Module Temperature
(Irradiation weighted)
• O&M activities as filter option
• Include required filters from IEC 61724 Ed.2 – 2917
Parameters (selection)
• Normalized parameters for direct comparison
• Weather types to compare similar weather, as energy
yield is influenced by several parameters at different
weather
• Real Nominal Power “Real Pnom” (extrapolation to STC,
PTC)

Technical Features (selection) II
Functions (selection)
5/18/2017
Analysis Functions
• Map view
• Dashboard
• Heatmaps
• Benchmark Performance of all installations
• Separation of main losses : Irradiation, Temperature,
Soiling/Dust, AOI
• Behavior vs. Time, Temperature, Irradiation, Season
• Long term degradation
• Extrapolation to STC, NOCT conditions
• Empirical data fits: compare trend with outliers
• Performance Guarantees: target vs. actual
• String level visualization (voltage & current)
Data handling
• Real time analysis platform
• Database cluster with highest availability (Fail over)
• Automated, decentralized backup
• Hosted according to EU regulations
• Accessible from mobile devices with HTML5, W3C
• Loss stages: The losses of the PV Power plant can be
analysed and checked vs. typical values.
• Different user level with selected rights

Conventions: Naming, Unit, Color
5/18/2017

Mechanistic Power Model (MPM)
5/18/2017

What is an empirical model?
• It’s a simple mathematical model for calculating
PMAX as a function of weather inputs
PMAX = GI*{C1*fn1(GI,TMOD…) + C2*fn2(GI,TMOD…) + … }
Constant Irradiance Module_Temp
Empirical Fit Coefficients Sum or product of terms 1..n
Input Dependencies
• It doesn't need any physical understanding
 it’s simple  values aren’t useful
• It should be able to be fitted by any simple software e.g. Excel solver (rather than specialised fitting software)
5/18/2017

How is an empirical model used?
• Determining bad measurement data (out of usual range)
• Interpolation of missing PMAX values
• Instantaneous performance validation
• Predicting performance at given conditions e.g. STC
• Simple energy yield estimation
Summing predicted PMAX vs. climate data (GI, TMOD …)
5/18/2017

We can improve models by normalising them and making them more
“Mechanistic”
Empirical Model
Not normalised. Coefficients scale with array size or module numbers
“Meaningless parameters” such as “TAMB*TMOD”
No idea what values mean good performance
e.g. PMEAS = GI * i=1..n Ci * fni(GI,TMOD…)
Mechanistic Model
Normalise coefficients by dividing by reference values e.g. nVOC = VOC.MEASURED/VOC.REFERENCE
Now we can more easily compare modules and understand degradation changes
e.g. PRDC= (PMEAS/PNOM/GI)= C1 + C2*Tmod + C3*Ln(Gi) + C4*Gi + C5*WS + ?
P TOLERANCE GAMMA LLEC RS WIND
% %/K %@LIC %@STC %/(ms-1)
5/18/2017

A simple normalised 6 parameter mechanistic model (L)
PRDC equation
5/18/2017
• The PRDC is the sum of each of these terms
• Plot on a stacked chart to determine the value of
each term and its shape vs. irradiance
• Some terms may be redundant or insignificant
e.g. C3 vs. C6
PRDC =
C1
+ C2*dTMOD
+ C3*ln(GI)
+ C4*GI
+ C5*WS
+ C6/GI
+ ...

A simple normalised 6 parameter mechanistic model (L)
PRDC equation
5/18/2017
PRDC =
C1
+ C2*dTMOD
+ C3*ln(GI)
+ C4*GI
+ C5*WS
+ C6/GI
+ ...
PRDC equation
PRDC vs. irradiance =
Sum +ve and -ve coefficients
How many terms are independent?
How many are significant?

PRDC vs. Irradiance for different technologies – Model L
PRDC = C1 + C2*dTMOD + C3*ln(GI) + C4*GI + C5*WS + C6/GI
[CONST] [ dTMOD ] [ ln(Gi) ] [ Gi ] [ WS ] [ 1/GI ]
5/18/2017
CdTe c-Si a-Si:uc-Si
Simple to fit Worst dTmod coeff Flattest PRDC vs Irradiance

Best fits to PRDC vs Gi c-Si for models A, C, D, E and J
vs. New Model L (Gantner Instruments data)
5/18/2017
“Normal conditions”
A’ C D
E J L
Empirical
Empirical
Empirical
Mechanistic Mechanistic
Mechanistic
New model L has sensible looking fit

PV Performance understanding
5/18/2017
Performanceunderstanding

Gantner Instruments Test Site, AZ/USA
Characterise modules outdoor where possible - to develop performance algorithms
5/18/2017
2 D Tracker:
6 channels w/
spectrometer
Weather sensors:
Irradiance,
Temperature, Wind,
Spectrum, RH, ..
24 Channels fixed: c-Si (ABC, HIT, n-type, ..), CdTe, CI(G)S, a-Si, a-Si:uc-Si; 6 tracked

Gantner Instruments Test Site, AZ/USA
Characterise modules outdoor where possible - to develop performance algorithms
5/18/2017
• Unique PV Module performance track record since 2010
• Baseline for next generation of PV Modeling and prediction of PV Plant
performance and monitoring
• Technology benchmark
• Bankability support for EPCs, Investors, Insurance
• Key for improved Utility PV Monitoring concepts
USP:
High performance data acquisition, 3rd party technology benchmark, advanced
characterization and Analysis methods
Tested PV Module references:
First Solar, Trina, Yingli, Sunpower, Manz, SolarWorld, Solar Frontier, Kaneka,
Hanergy, MiaSole, Q-Cells, Bosch Solar, Panasonic
Cooperations with leading Institutions, experts, EPCs and IPPs.
Benefits

Loss Factor Model (LFM) – used by leading institutes, PVLib, ..
LFM
• 6 normalised orthogonal parameters associated with
ISC, RSC, IMP, VMP, ROC and VOC
• Multiply all 6 together to give the PRDC or MPR
• Easy sanity check for bad data if not ~1
• Easy to see good from bad values e.g. 98% or 90%
• Differentiates PV technologies –
e.g. Thin Film usually has a higher loss due to ROC
(from TCO resistivity)
Use to characterise PV Modules based on IV measurements
5/18/2017
Plot LFM values vs To show
Time Degradation or annealing
Irradiance Low light performance
Module Temperature Thermal coefficients

Gantner.webportal
Demo Links / Screenshots
5/18/2017

Structure
5/18/2017

5/18/2017

Aggregation
5/18/2017

5/18/2017

Normalization
5/18/2017

5/18/2017

Correction
5/18/2017

5/18/2017

Benchmark
5/18/2017

5/18/2017

Mechanistic Power Model
5/18/2017

5/18/2017

21 gi 8th_pvpmc_talk_abq_final_ext_170508t19_new

Recommandé

Recommandé

Contenu connexe

Tendances

Tendances (20)

Similaire à 21 gi 8th_pvpmc_talk_abq_final_ext_170508t19_new

Similaire à 21 gi 8th_pvpmc_talk_abq_final_ext_170508t19_new (20)

Plus de Sandia National Laboratories: Energy & Climate: Renewables

Plus de Sandia National Laboratories: Energy & Climate: Renewables (20)

Dernier

Dernier (20)

21 gi 8th_pvpmc_talk_abq_final_ext_170508t19_new