NREL is a national laboratory of the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, operated by the Alliance for Sustainable Energy, LLC.
Next-Generation Satellite Modeling for the
National Solar Radiation Database
(NSRDB)
Dr. Manajit Sengupta
Aron Habte , Anthony Lopez, and Andrew Weekley, NREL
Christine Molling CIMMS, University of Wisconsin
Andrew Heidinger, NOAA
PV Modeling Workshop, Cologne, Germany
October 22-23, 2015
Work is funded by the US Department of Energy
2
Evolution of Solar Resource Data
1952-1975 SOLMET1 [ERDA, NOAA, 1979]
1961-1990 NSRDB2 [DOE, NOAA, 1994]
1991-2005 NSRDB-II3 [DOE, NOAA, 2007]
1998-2014 NSRDB [DOE, NOAA, UW 2015]
National Solar Radiation Data Base
(1)
248 stations with
26 Measurement
Stations
1977-80
(2)
239
Modeled
Stations with
56 partial
measureme
nt stations
1990
(3)
1,454 Modeled
Locations
1991-2005
http://nsrdb.nrel.gov
Satellite-based, gridded
4 km x 4 km
Half-hourly
1998-2014
3
• Empirical Approach (Industry standard
traditional approach):
– Build model relating satellite measurements and ground
observations.(cloud index and clearness index)
– Use those models to obtain solar radiation at the surface
from satellite measurements.
• Physical Approach: (the new approach)
– Retrieve cloud and aerosol information from satellites
– Use the information in a radiative transfer model
How do satellites model surface radiation?
4
F+
G = a – b F-
TOA
F+
TOA F+
TOA
Basic principle
Richard Perez, et al.
Clearness Index Satellite Reflectance
(cloud Index)
Empirical Approach to Satellite Modeling
5
Satellite image Cloud Properties
Solar Radiation
Satellite based
Cloud Retrieval
Model
Radiative
Transfer Models
Physical Approach to Satellite Modeling
6
Physical Approach to Satellite Modeling
Satellite Retrieval:
Inputs: Satellite radiance from 5
channels of GOES
Outputs: cloud mask
Cloud properties including cloud type
and cloud optical thickness
Ancillary data:
Aerosols from MODIS and MISR satellites
Ozone TOMS/OMI satellites
Water Vapor from NWP model (GFS)
Snow from NSIDC
Atmospheric Profiles Temperature,
Pressure from (GFS)
Radiative Transfer Model:
Clear Sky: REST2
All Sky: FARMS
Inputs: Aerosol, Water Vapor, Ozone,
Elevation, cloud mask, cloud properties
Output: GHI and DNI under all sky conditions
7
GIS based data access with web-service for multi-pixel download
Includes ancillary meteorological data for PV/CST modeling using SAM
Accessing the NSRDB Data
http://nsrdb.nrel.gov
8
Product Timeline
• Beta product (2005-2012) currently
online (V. 1)
• V2 Product (1998-2014) available by
October 2015
• Typical Meteorological Year (TMY)
product by October 2015
• Quarterly monthly update available
from 2016
• From 2016 annual datasets will be
available by following March
9
Validation of Satellite product (V1.0.1) using Ground Data
http://www.esrl.noaa.gov/gmd/grad/surfrad/
Code Name Latitude Longitude Elevation Time Zone Installed
BND Bondville, Illinois 40.05° N 88.37° W 230 m 6 hours from UTC Apr-94
TBL Table Mountain, Boulder, Colorado 40.13° N 105.24° W 1689 m 7 hours from UTC Jul-95
DRA Desert Rock, Nevada 36.63° N 116.02° W 1007 m 8 hours from UTC Mar-98
FPK Fort Peck, Montana 48.31° N 105.10° W 634 m 7 hours from UTC Nov-94
GCM Goodwin Creek, Mississippi 34.25° N 89.87° W 98 m 6 hours from UTC Dec-94
PSU Penn. State Univ., Pennsylvania 40.72° N 77.93° W 376 m 5 hours from UTC Jun-98
SXF Sioux Falls, South Dakota 43.73° N 96.62° W 473 m 6 hours from UTC Jun-03
NOAA SURFRAD
DATASTREAMS:
GHI, DNI and Diffuse
Comparison with 2005-2012
satellite product
10
Validation with Surface Measurements
GHI: Mean Bias Error (MBE)
±5%
11
Validation with Surface Measurements
GHI: Root Mean Square Deviation
(RMSD)
12
Validation with Surface Measurements
DNI: Mean Bias Error (MBE)
±5%
13
Validation with Surface Measurements
DNI: Root Mean Square Deviation
(RMSD)
14
Validation with Surface Measurements
Bondville, IL: GHI for all cases
15
Bondville, IL: GHIClear Sky Cases
Cloudy Sky Cases
Validation with Surface Measurements
16
Conclusions
• New gridded satellite product available publicly
from NREL (http://nsrdb.nrel.gov )
• Datasets is 4 km, 30 minute resolution with
meteorological variable from NASA MERRA.
• 2005-2012 currently available with 1998-2014
online by the end of October.
• Accurate aerosol and water vapor information is
critical in properly modeling clear sky GHI and
DNI.
• Improved All Sky model FARMS for cloudy sky.
• Significant uncertainty in cloudy cases.
17
Future Work
• Inclusion of daily variability of aerosols from
MACC/aerosols.
• Improved surface albedo time series to reflect land
use changes.
• Improved identification of high albedo surfaces
(sand and snow).
• 5 minute data from GOES-R.
• Spectral long-term datasets in Plane of Array.
18
Thank You! Contact: manajit@nrel.gov
http://www.nrel.gov/solar_radiation
http://www.nrel.gov/docs/fy15osti/63112.pdf
http://nsrdb.nrel.gov

15 sengupta next_generation_satellite_modelling

  • 1.
    NREL is anational laboratory of the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, operated by the Alliance for Sustainable Energy, LLC. Next-Generation Satellite Modeling for the National Solar Radiation Database (NSRDB) Dr. Manajit Sengupta Aron Habte , Anthony Lopez, and Andrew Weekley, NREL Christine Molling CIMMS, University of Wisconsin Andrew Heidinger, NOAA PV Modeling Workshop, Cologne, Germany October 22-23, 2015 Work is funded by the US Department of Energy
  • 2.
    2 Evolution of SolarResource Data 1952-1975 SOLMET1 [ERDA, NOAA, 1979] 1961-1990 NSRDB2 [DOE, NOAA, 1994] 1991-2005 NSRDB-II3 [DOE, NOAA, 2007] 1998-2014 NSRDB [DOE, NOAA, UW 2015] National Solar Radiation Data Base (1) 248 stations with 26 Measurement Stations 1977-80 (2) 239 Modeled Stations with 56 partial measureme nt stations 1990 (3) 1,454 Modeled Locations 1991-2005 http://nsrdb.nrel.gov Satellite-based, gridded 4 km x 4 km Half-hourly 1998-2014
  • 3.
    3 • Empirical Approach(Industry standard traditional approach): – Build model relating satellite measurements and ground observations.(cloud index and clearness index) – Use those models to obtain solar radiation at the surface from satellite measurements. • Physical Approach: (the new approach) – Retrieve cloud and aerosol information from satellites – Use the information in a radiative transfer model How do satellites model surface radiation?
  • 4.
    4 F+ G = a– b F- TOA F+ TOA F+ TOA Basic principle Richard Perez, et al. Clearness Index Satellite Reflectance (cloud Index) Empirical Approach to Satellite Modeling
  • 5.
    5 Satellite image CloudProperties Solar Radiation Satellite based Cloud Retrieval Model Radiative Transfer Models Physical Approach to Satellite Modeling
  • 6.
    6 Physical Approach toSatellite Modeling Satellite Retrieval: Inputs: Satellite radiance from 5 channels of GOES Outputs: cloud mask Cloud properties including cloud type and cloud optical thickness Ancillary data: Aerosols from MODIS and MISR satellites Ozone TOMS/OMI satellites Water Vapor from NWP model (GFS) Snow from NSIDC Atmospheric Profiles Temperature, Pressure from (GFS) Radiative Transfer Model: Clear Sky: REST2 All Sky: FARMS Inputs: Aerosol, Water Vapor, Ozone, Elevation, cloud mask, cloud properties Output: GHI and DNI under all sky conditions
  • 7.
    7 GIS based dataaccess with web-service for multi-pixel download Includes ancillary meteorological data for PV/CST modeling using SAM Accessing the NSRDB Data http://nsrdb.nrel.gov
  • 8.
    8 Product Timeline • Betaproduct (2005-2012) currently online (V. 1) • V2 Product (1998-2014) available by October 2015 • Typical Meteorological Year (TMY) product by October 2015 • Quarterly monthly update available from 2016 • From 2016 annual datasets will be available by following March
  • 9.
    9 Validation of Satelliteproduct (V1.0.1) using Ground Data http://www.esrl.noaa.gov/gmd/grad/surfrad/ Code Name Latitude Longitude Elevation Time Zone Installed BND Bondville, Illinois 40.05° N 88.37° W 230 m 6 hours from UTC Apr-94 TBL Table Mountain, Boulder, Colorado 40.13° N 105.24° W 1689 m 7 hours from UTC Jul-95 DRA Desert Rock, Nevada 36.63° N 116.02° W 1007 m 8 hours from UTC Mar-98 FPK Fort Peck, Montana 48.31° N 105.10° W 634 m 7 hours from UTC Nov-94 GCM Goodwin Creek, Mississippi 34.25° N 89.87° W 98 m 6 hours from UTC Dec-94 PSU Penn. State Univ., Pennsylvania 40.72° N 77.93° W 376 m 5 hours from UTC Jun-98 SXF Sioux Falls, South Dakota 43.73° N 96.62° W 473 m 6 hours from UTC Jun-03 NOAA SURFRAD DATASTREAMS: GHI, DNI and Diffuse Comparison with 2005-2012 satellite product
  • 10.
    10 Validation with SurfaceMeasurements GHI: Mean Bias Error (MBE) ±5%
  • 11.
    11 Validation with SurfaceMeasurements GHI: Root Mean Square Deviation (RMSD)
  • 12.
    12 Validation with SurfaceMeasurements DNI: Mean Bias Error (MBE) ±5%
  • 13.
    13 Validation with SurfaceMeasurements DNI: Root Mean Square Deviation (RMSD)
  • 14.
    14 Validation with SurfaceMeasurements Bondville, IL: GHI for all cases
  • 15.
    15 Bondville, IL: GHIClearSky Cases Cloudy Sky Cases Validation with Surface Measurements
  • 16.
    16 Conclusions • New griddedsatellite product available publicly from NREL (http://nsrdb.nrel.gov ) • Datasets is 4 km, 30 minute resolution with meteorological variable from NASA MERRA. • 2005-2012 currently available with 1998-2014 online by the end of October. • Accurate aerosol and water vapor information is critical in properly modeling clear sky GHI and DNI. • Improved All Sky model FARMS for cloudy sky. • Significant uncertainty in cloudy cases.
  • 17.
    17 Future Work • Inclusionof daily variability of aerosols from MACC/aerosols. • Improved surface albedo time series to reflect land use changes. • Improved identification of high albedo surfaces (sand and snow). • 5 minute data from GOES-R. • Spectral long-term datasets in Plane of Array.
  • 18.
    18 Thank You! Contact:manajit@nrel.gov http://www.nrel.gov/solar_radiation http://www.nrel.gov/docs/fy15osti/63112.pdf http://nsrdb.nrel.gov