ERCOT Solar Modeling – ETWG Meeting April 30, 2013

Project Overview ▫ Objective: Produce 20 years of high-quality hourly solar production data for several different solar technologies in all 254 Texas counties ▫ Methodology: Utilize National Solar Radiation Database (NSRDB) Class I and Class II weather station data and solar production modeling software to generate 4,452 individual datasets with 8,760 points each

Agenda ▫ Origin and quality of raw data ▫ How solar production modeling software works ▫ Weather station-county grouping ▫ Summary of results ▫ Quality assurance and statistical analysis ▫ Q&A

Project Overview
Task 1: Collect Data
• Collected 20 years (1991-2010) of Class I, II & III weather data and TMY3 data for 89 stations across Texas and 4 stations from adjacent states (53 Class I / II) •Converted data into format compatible with NREL SAM modeling software
Task 2: Develop Models
•Evaluated typical configurations for four different technology types Residential (5 kW), CSP (50MW), SAT (1MW) and Fixed tilt crystalline (1MW) given Texas solar radiation and geography. •Developed technology-specific models in NREL System Advisor Model (SAM) software using typical configurations.
Task 3: Run Simulations
•Assigned groups of counties to Class I/II weather stations using NREL “areas of influence” algorithm and data.
•Ran model simulations for each of the four technologies using relevant historical NSRDB and TMY data from the 34 weather stations selected. Simulations were run for the remaining weather stations as well.
Task 4: Sensitivity and Uncertainty Analysis
•Phase I – Data Quality Assessment: Variance analysis to identify outlier or anomalous data points in model results •Phase II – Sensitivity Analysis: Present effects of geographic distribution and technology type on model results.
Task 5: Report and Presentation
•Final deliverables: (1) all data collected/generated in .ZIP file (2.8GB); (2) summary data files containing only AC output model results for each of the 34 weather stations and four technology type; (3) Summary report and final presentation.

Project Overview
Task 1: Collect Data
• Collected 20 years (1991-2010) of Class I, II & III weather data and TMY3 data for 89 stations across Texas and 4 stations from adjacent states (53 Class I / II) •Converted data into format compatible with NREL SAM modeling software
Task 2: Develop Models
•Evaluated typical configurations for four different technology types Residential (5 kW), CSP (50MW), SAT (1MW) and Fixed tilt crystalline (1MW) given Texas solar radiation and geography. •Developed technology-specific models in NREL System Advisor Model (SAM) software using typical configurations.
Task 3: Run Simulations
•Assigned groups of counties to Class I/II weather stations using NREL “areas of influence” algorithm and data.
•Ran model simulations for each of the four technologies using relevant historical NSRDB and TMY data from the 34 weather stations selected. Simulations were run for the remaining weather stations as well.
Task 4: Sensitivity and Uncertainty Analysis
•Phase I – Data Quality Assessment: Variance analysis to identify outlier or anomalous data points in model results •Phase II – Sensitivity Analysis: Present effects of geographic distribution and technology type on model results.
Task 5: Report and Presentation
•Final deliverables: (1) all data collected/generated in .ZIP file (2.8GB); (2) summary data files containing only AC output model results for each of the 34 weather stations and four technology type; (3) Summary report and final presentation.

Task 1: Collect Data Meteorological Data Required for Solar Simulations
Entry Dry-bulb temperature Dew-point temperature Relative humidity Wind speed Wind direction Atmospheric pressure Global horizontal radiation Direct normal radiation Diffuse horizontal radiation Units
°C °C % m/s deg mbar W/m² W/m² W/m² System Advisory Model (SAM) Weather Data Documentation- 12/7/2011

National Solar Radiation Database (NSRDB) ▫ 20 years (1991-2010) of measured/modeled solar radiation data with accompanying meteorological fields for weather stations across the US ▫ Solar radiation data generated by the NREL Meteorological-Statistical Model (METSTAT) and a State University of New York (SUNY) developed model based on satellite data ▫ Weather data from National Climate Data Center (NCDC)

NSRDB ▫ Four levels of data-filling methods – Short-term interpolation - up to 5 hour gaps and gaps at night – Medium-term filling - gaps up to 24 hours – Long-term filling - gaps up to 1 year – Last-ditch filling - gaps greater than a year – Level of data filling required determines Class of
weather station

NSRDB ▫ Class I Stations – Complete period of record (all hours 1991–2010) for solar and key meteorological fields – Highest-quality solar modeled data (16 sites in Texas) ▫ Class II Stations – Complete period of record but significant periods of interpolated, filled, or otherwise lower-quality input data for the solar models (37 in Texas).
▫ Class III Stations – Have some gaps in the period of record but have at least 3 years of data that might be useful for some applications (36 in Texas) National Solar Radiation Database (NSRDB) 1991-2010 User’s Manual

Class I / II Weather Stations in TX 53 Class I/II NSRDB Weather Stations in TX Red = Class I White = Class II URS GIS Dept.

Raw Data for this Project
NSRDB Meteorological Data
• 20 years (1991-2010) of hourly meteorological and solar radiation data for 53 weather stations in and around Texas, including: – Dry-bulb temperature – Dew-point temperature – Relative humidity – Wind speed – Wind direction – Atmospheric pressure TMY Algorithm Statistical analysis selects 12 “typical” months from 20-year data set to complete one full year – Global horizontal radiation – Direct normal radiation – Diffuse horizontal radiation
Typical Meteorological Data (TMY)
• TMY files for 53 weather stations (Class I & II) • Composed of 12 typical meteorological months (January through December) that are concatenated to form a single year.
• One TMY file contains a single representative year (8,760 data points).

Data Conversion ▫ NSRDB data conversion process – National Renewable Energy Laboratory (NREL) System Advisor Model (SAM) requires weather data in TMY2, TMY3 or EPW format – URS converted NCDC-formatted NSRDB data to the TMY3 format for use as an input to SAM. The conversion process was automated using Bash shell scripts and Unix programming languages, including AWK and SED.
– Result was 21 TMY3-formatted meteorological data
files for each of the 53 Class I/II weather stations (1,113 TMY3 files total)

Project Overview
Task 1: Collect Data
• Collected 20 years (1991-2010) of Class I, II & III weather data and TMY3 data for 89 stations across Texas and 4 stations from adjacent states (53 Class I / II) •Converted data into format compatible with NREL SAM modeling software
Task 2: Develop Models
•Evaluated typical configurations for four different technology types Residential (5 kW), CSP (50MW), SAT (1MW) and Fixed tilt crystalline (1MW) given Texas solar radiation and geography. •Developed technology-specific models in NREL System Advisor Model (SAM) software using typical configurations.
Task 3: Run Simulations
•Assigned groups of counties to Class I/II weather stations using NREL “areas of influence” algorithm and data.
•Ran model simulations for each of the four technologies using relevant historical NSRDB and TMY data from the 34 weather stations selected. Simulations were run for the remaining weather stations as well.
Task 4: Sensitivity and Uncertainty Analysis
•Phase I – Data Quality Assessment: Variance analysis to identify outlier or anomalous data points in model results •Phase II – Sensitivity Analysis: Present effects of geographic distribution and technology type on model results.
Task 5: Report and Presentation
•Final deliverables: (1) all data collected/generated in .ZIP file (2.8GB); (2) summary data files containing only AC output model results for each of the 34 weather stations and four technology type; (3) Summary report and final presentation.

National Renewable Energy Laboratory (NREL) System Advisor Model (SAM)

NREL SAM ▫ Available for free at NREL SAM website: – https://sam.nrel.gov/ – URS used most recent version (2013.1.15) – Selected for many reasons over other options: • Transparency/replicability of study results • Flexibility in raw data sourcing • Ability to run automated scripts for fast data processing

Solar Technologies ▫ Fixed tilt crystalline silicon ▫ Single-axis tracking (SAT) ▫ Residential rooftop (fixed C-Si) ▫ Concentrating Solar Power (CSP)

Fixed Tilt Crystalline Silicon (C-Si) ▫ Majority of PV arrays composed of C-Si cells ▫ Lowest initial cost per watt ▫ Polycrystalline efficiency: 13-16% ▫ Monocrystalline efficiency: 14-20% ▫ Ideally oriented due south with slope (tilt) of modules equal to latitude

Single-Axis Tracker (SAT) ▫ Increase the energy produced by a solar array by tracking the sun from east to west diurnally ▫ Tilt of SAT systems are usually kept flat, normal to the zenith, and have a total east-west tracking range of 90°.
▫ SAT systems are designed to maximize land use without causing self-shading

Rooftop Residential ▫ Utilize the same components of a fixed-tilt system, but is constrained to the orientation of the building roof ▫ Typically range from 2kW to 16kW

Concentrating solar power (CSP) ▫ Utilize concentrated solar radiation to generate thermal energy, which is used to power a conventional electricity producing steam turbine generator ▫ Two primary types of plants: parabolic troughs and power towers ▫ URS elected to model a parabolic trough type CSP plant

Solar Radiation Levels across US

CSP Resource in Texas

Final Datasets ▫ Individual models configured in SAM for each technology type – Fixed tilt C-Si = 1MW – SAT = 1MW – Residential = 10kW – CSP = 50MW ▫ 20° selected for slope of fixed tilt system – most typical installation in Texas

Project Overview
Task 1: Collect Data
• Collected 20 years (1991-2010) of Class I & II weather data and TMY3 data for 53 stations across Texas •Converted data into format compatible with NREL SAM modeling software
Task 2: Develop Models
•Evaluated typical configurations for four different technology types Residential (5 kW), CSP (50MW), SAT (1MW) and Fixed tilt crystalline (1MW) given Texas solar radiation and geography. •Developed technology-specific models in NREL System Advisor Model (SAM) software using typical configurations.
Task 3: Run Simulations
•Assigned groups of counties to Class I/II weather stations using NREL “areas of influence” algorithm and data.
•Ran model simulations for each of the four technologies using relevant TMY data from the 34 weather stations selected.
Task 4: Sensitivity and Uncertainty Analysis
•Phase I – Data Quality Assessment: Variance analysis to identify outlier or anomalous data points in model results •Phase II – Sensitivity Analysis: Measure effects of geographic distribution, orientation of collectors, and technology type on model results •Phase III – Uncertainty Analysis: Calculate uncertainty caused by key measurement variables using statistical analysis.
Task 5: Report and Presentation
•Final deliverables: (1) all data collected/generated in .ZIP file (2.8GB); (2) summary data files containing only AC output model results for each of the 34 weather stations and four technology type; (3) Summary report and final presentation.

Station-County Grouping ▫ US DOE used algorithms to define “zones of influence” around Class I/II weather stations at 10km 2 blocks

Station-County Grouping

Station-County Grouping

Final Model Result Datasets ▫ Each of 1,113 TMY3 files fed into each model using automated script ▫ Output was 4,452 individual solar production model results containing 8,760 rows ▫ Post-production of data to generate total hourly AC output zeroing out any overnight consumption ▫ Results for each of four technologies available for any county in Texas based on associated weather station

20-Year Production Model Results
20-Year Annual Electricity Production Model Results Lufkin Angelina Co (722446)
160,000 140,000 120,000 100,000 80,000 60,000 40,000 20,000 0 1991 1992 1993 1994 1995 1996 1997 1998 1999 PVSAT 2000 PVFT RES 2001 CSP 2002 2003 2004 2005 2006 2007 2008 2009 2010

20-Year Production Model Results
20-Year Annual Electricity Production Model Results Austin Mueller Muni AP (UT) (722540)
160,000 140,000 120,000 100,000 80,000 60,000 40,000 20,000 0 1991 1992 1993 1994 1995 1996 1997 1998 1999 PVSAT 2000 PVFT RES 2001 CSP 2002 2003 2004 2005 2006 2007 2008 2009 2010

20-Year Production Model Results
20-Year Annual Electricity Production Model Results Brownsville/So Padre Island Intl AP (722500)
160,000 140,000 120,000 100,000 80,000 60,000 40,000 20,000 0 1991 1992 1993 1994 1995 1996 1997 1998 1999 PVSAT 2000 PVFT RES 2001 CSP 2002 2003 2004 2005 2006 2007 2008 2009 2010 Model results normalized to 50 MW to match CSP system

20-Year Production Model Results
20-Year Annual Electricity Production Model Results Dallas Ft. Worth Intl AP (722590)
160,000 140,000 120,000 100,000 80,000 60,000 40,000 20,000 0 1991 1992 1993 1994 1995 1996 1997 1998 1999 PVSAT 2000 PVFT RES 2001 CSP 2002 2003 2004 2005 2006 2007 2008 2009 2010

20-Year Production Model Results
20-Year Annual Electricity Production Model Results Lubbock Intl AP (722670)
160,000 140,000 120,000 100,000 80,000 60,000 40,000 20,000 0 1991 1992 1993 1994 1995 1996 1997 1998 1999 PVSAT 2000 PVFT RES 2001 CSP 2002 2003 2004 2005 2006 2007 2008 2009 2010

20-Year Production Model Results
20-Year Annual Electricity Production Model Results El Paso Intl AP (722700)
160,000 140,000 120,000 100,000 80,000 60,000 40,000 20,000 0 1991 1992 1993 1994 1995 1996 1997 1998 1999 PVSAT 2000 PVFT RES 2001 CSP 2002 2003 2004 2005 2006 2007 2008 2009 2010

Project Overview
Task 1: Collect Data
• Collected 20 years (1991-2010) of Class I, II & III weather data and TMY3 data for 89 stations across Texas and 4 stations from adjacent states (53 Class I / II) •Converted data into format compatible with NREL SAM modeling software
Task 2: Develop Models
•Evaluated typical configurations for four different technology types Residential (5 kW), CSP (50MW), SAT (1MW) and Fixed tilt crystalline (1MW) given Texas solar radiation and geography. •Developed technology-specific models in NREL System Advisor Model (SAM) software using typical configurations.
Task 3: Run Simulations
•Assigned groups of counties to Class I/II weather stations using NREL “areas of influence” algorithm and data.
•Ran model simulations for each of the four technologies using relevant historical NSRDB and TMY data from the 34 weather stations selected. Simulations were run for the remaining weather stations as well.
Task 4: Sensitivity and Uncertainty Analysis
•Phase I – Data Quality Assessment: Variance analysis to identify outlier or anomalous data points in model results •Phase II – Sensitivity Analysis: Present effects of geographic distribution and technology type on model results.
Task 5: Report and Presentation
•Final deliverables: (1) all data collected/generated in .ZIP file (2.8GB); (2) summary data files containing only AC output model results for each of the 34 weather stations and four technology type; (3) Summary report and final presentation.

Data Quality & Statistical Analysis ▫ Final datasets reviewed for data quality and alignment with typical production profiles using URS internal technical review (ITR) process 1000 900 800 700 600 500 400 300 200 100 0 Austin Mueller production model results – PVFT (blue), SAT (red) – 7/1/10 – 7/8/10

Statistical Variability Calculations ▫ Purpose: calculate the likelihood that a solar plant will generate a certain amount of electricity in any given year ▫ Exceedance probabilities (P50 & P90) can be calculated by fitting the dataset to a standard probability distribution and determining value from the cumulative distribution function (CDF).
▫ P50/P90 calculations prepared for six stations across TX using 20 years of production model results

Exceedance Probabilities
Lubbock Intl AP (722670)
1.000
0.900
0.800
0.700
0.600
0.500
0.400
0.300
0.200
0.100
0.000
2,000 2,050 2,100 2,150 2,200
MWh
2,250 2,300 2,350 2,400 1.000
0.900
0.800
0.700
0.600
0.500
0.400
0.300
0.200
0.100
0.000
1,600 1,700
Dallas Ft Worth Intl AP (722590)
1,800
MWh
1,900 2,000
Austin Mueller Muni AP (UT) (722540)
2,100 1.000
0.900
0.800
0.700
0.600
0.500
0.400
0.300
0.200
0.100
0.000
1,600 1,700 1,800
MWh
1,900 2,000 2,100 1.000
0.900
0.800
0.700
0.600
0.500
0.400
0.300
0.200
0.100
0.000
2,100 2,200
El Paso Intl AP (722700)
2,300
MWh
2,400 2,500 1.000
0.900
0.800
0.700
0.600
0.500
0.400
0.300
0.200
0.100
0.000
1,600 1,700
Lufkin Angelina Co (722446)
1,800
MWh
1,900 2,000 2,100 1.000
0.900
0.800
0.700
0.600
0.500
0.400
0.300
0.200
0.100
0.000
1,600
Brownsville/So Padre Intl AP (722500)
1,700 1,800
MWh
1,900 2,000 2,100

Weather Station Variance Analysis ▫ Graph below shows the % variance in total annual production model estimates between the Class I station shown and a nearby Class I or II station ▫ Many Class II stations relied upon inferior statistically derived cloud cover data prior to 1998 when satellite imagery data was introduced 15.0% 10.0% 5.0% 0.0% -5.0% -10.0% -15.0% 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 DFW Intl Airport Houston Bush Intl AP San Antonio Intl AP Austin Mueller AP (UT) Lubbock Intl AP

Angle of Incidence and SAT ▫ During winter months, SAT systems have “dip” in production during middle of day due to angle of incidence 900 800 700 600 500 400 300 200 100 0 SAT production model results - Odessa (blue), Nacogdoches (red) – 1/1/10 – 1/2/10

Angle of Incidence and SAT Black & Veatch Solar Diagnostics Presentation

2011 Data ▫ Kevin contacted Michael Milligan at NREL to ask about availability of 2011 NCDC format NSRDB hourly solar data with the filled meteorological fields for the Class I & II weather stations in Texas ▫ No reliable 2011 meteorological data acquired in time for this project

Q&A – Mike Hoffman, Project Manager mike.hoffman@urs.com
W: (512) 419-6092 C: (512) 297-5116