Uncertainty of the Solargis solar radiation database
-
Upload
sandia-national-laboratories-energy-climate-renewables -
Category
Technology
-
view
211 -
download
2
Transcript of Uncertainty of the Solargis solar radiation database
Uncertainty of the Solargis solar radiation database
PV Performance Modeling and Monitoring WorkshopFreiburg, 24 Oct 2016
Marcel Suri, Tomas Cebecauer, Jose A. Ruiz-Arias and Juraj Betak
Solargis, Slovakia
PV Performance Modeling and Monitoring Workshop, Freiburg, 24 Oct 2016 2
About Solargis
Solar resource, meteorological and photovoltaic simulation data, software and expert services
• Prospection
• Project development
• Monitoring
• Forecasting
historical and recent
forecast NWP modelnowcast
PV Performance Modeling and Monitoring Workshop, Freiburg, 24 Oct 2016 3
Solargis solar resource database
GHI
DNI
PV Performance Modeling and Monitoring Workshop, Freiburg, 24 Oct 2016 4
Continuous geographical coverage
Spatial and temporal consistency
High resolution
High availability
History of 10 to 22+ years
Systematic update for monitoring and nowcasting
Historical and operational satellite-based data
Data inputs: JMA, ECMWF, NOAA, SRTM
Source: Solargis
Difference model - measurements
Factors that determine the difference between the model and measurements
Models• Mathematical and algorithmic formulation of models• Input data sets (satellite, weather models, etc.)
Solar monitoring instruments*• Accuracy of sensors• Maintenance and calibration of the instruments• Quality control of the measured data
*For validation, in major cases, we use high quality measurements
PV Performance Modeling and Monitoring Workshop, Freiburg, 24 Oct 2016 6
Solargis uncertainty of yearly estimates
GHI: ±4% to ±8%
±3.9%**
±7.6%**
* 68.27% occurrence: standard deviation (STDEV) assuming simplified assumption of normal distribution** 80% occurrence: calculated as 1.28155 STDEV − can be used for an estimate of P90 values
DNI: ±8% to ±15%
Source: SolarGIS
PV Performance Modeling and Monitoring Workshop, Freiburg, 24 Oct 2016 7
Solargis uncertainty: focus on yearly GHI
Model deviation yearly GHI
Distance to validation sites
GHI: ±4% to ±8%
PV Performance Modeling and Monitoring Workshop, Freiburg, 24 Oct 2016 8
Solargis model
PV Performance Modeling and Monitoring Workshop, Freiburg, 24 Oct 2016 9
Solargis satellite model
Atmospheric data Satellite dataTerrain data
PV Performance Modeling and Monitoring Workshop, Freiburg, 24 Oct 2016 10
Solar radiation modelInputs and outputs
Solargis inputs(Americas)
Source of input data
Time representation
Original time step
Approx. grid resolution
Cloud indexGOES
(NOAA)1999 to date 30 minutes 3 km
Atmospheric Optical Depth (aerosols)
MACC-II(ECMWF)
2003 to date6 hours
(monthly until 2002)85 and 125 km
Water vaporCFSR/GFS(NOAA)
1999 to date 1 and 3 hours 35 and 55 km
Elevation and horizonSRTM-3(SRTM)
- - 250 m
Solargis primary data outputs (GHI and DNI)
- 1999 to date 30 minutes 250 m
Source: Solargis
Example region: Peru
Errors of model values
The total error of the model values is a combination of:
1. Errors associated to the representation of solar radiation with physical formulation (the model)
2. Errors related to the model inputs: satellite data features, clouds, aerosols, water vapor, etc.
Methodology: Simplified framework
Total uncertainty σε2 is evaluated as:
• σm2 represents the background uncertainty of the model (constant)
• Model inputs xi are referred to as uncertainty drivers
• Φi(xi) are sensitivity functions determining how much each model input xi
contributes to the total uncertainty
Methodology: evaluation of the uncertainty driver
Satellite pixel distortion
For each driver the following is evaluated:
• Uncertainty model
• Distribution of values
Methodology: Uncertainty drivers
The uncertainty drivers:
• Clouds
• Aerosol optical depth
• Total water vapor
• Snow coverage
• Terrain variability
• Distance to water surface
• Anthropogenic pollution
• Satellite pixel distortion
For each uncertainty driver, the corresponding sensitivity function has been empirically calculated
Uncertainty drivers: Clouds
Computed as as P99(GHI)/GHICS
Key issue: high variability and persistence of clouds in tropics
Uncertainty drivers: Aerosol optical depth
Computed from weather reanalysis as AOD550 x AODerror*
Key issue: High atmospheric turbidity and high variability
* Ruiz-Arias et al. 2013, doi: 10.5194/acp-13-675-2013
Uncertainty drivers: Water vapour
Precipitable water computed from weather reanalysis data
Key issue: high precipitable water in equatorial tropics
Uncertainty drivers: Snow coverage
Computed from snow depth water equivalent (weather reanalysis)
Key issue: Ability of satellites to identify snow/ice/clouds
Uncertainty drivers: Terrain
Terrain variability computed as standard deviation of elevation from DEM
Key issue: Mountains with complex and fast changing patterns of clouds, shadows, elevation and albedo
Uncertainty drivers: Distance to water bodies
Computed as/from: GIS layers representing water bodies
Key issues: mixed pixels on the overlap of land and water vs. geometric instability of satellites
Uncertainty drivers: Anthropogenic pollution
Estimated from total SO2 emissions from EDGAR HTAP database
Key issue: Attenuation in high polluted areas
Uncertainty drivers: Satellite pixel distortion
Computed as change of pixel diagonal length
Shows magnitude of errors due to geometrical distortions in satellite data
Results: Total GHI uncertaintyLong-term value, in %
Background error = 3%
Comparing evidence with the uncertainty estimates
Comparing model with measurements
Factors that determine the difference between the model and measurements
Models• Mathematical and algorithmic formulation of models• Input data sets (satellite, weather models, etc.)
Solar monitoring instruments*• Accuracy of sensors• Maintenance and calibration of the instruments• Quality control of the measured data
PV Performance Modeling and Monitoring Workshop, Freiburg, 24 Oct 2016 26
Reducing uncertainty of yearly estimates by ground measurements
DNI: ±3.5%
GHI: ±2.5%
Best achievable uncertainty
0
2
4
6
8
10
12
14
16
0 12 24 36 48
Ach
ieva
ble
un
ce
rta
inty
(%
)
Period of ground measurements (months)
DNI P99.5
DNI P90
GHI P99.5
GHI P90
Running ground-monitoring campaign and combining measurements with satellite data is the way to maintain low uncertainty of solar resource data in a longterm
Values are indicative
Next steps
• Further reduction of uncertainty expected with new empirical evidence
• Results will be incorporated into the uncertainty processing chain
• Uncertainty map for DNI under development
• New uncertainty drivers and time aggregations will be considered