Cloud Cover Analysis
Matching 1-3 Hour Solar Radiation Forecast
figure 3: LAPS system. the Local Analysis and Prediction System (LAPS) is an example of how solar forecasts can be made.
Left panel, clouds are initialized at time zero of the model forecast and continue to progress through the first few hours of the
forecast (scale is 0 to 1). Right panel, a Weather Research and forecast (WRf) model solar radiation two-hour forecast that was
initialized using the LAPS cloud analysis. The global horizontal irradiance scale is 0−600 watts per square meter. Two hours into
the forecast, the solar radiation field closely matches the analyzed cloud cover locations from the same time on the left panel.
Any effort to improve forecasts of solar
resources will benefit from more observations
of solar irradiance from ground-based and satel-
lite-based sensors and improved modeling tech-
niques based on these observations. For exam-
ple, improvements in radiative transfer models,
which indicate how much solar radiation is lost
from scattering and absorption in a cloud-free
sky, can improve the quality of solar forecasts.
PV uses the total (diffuse plus direct) solar
irradiance falling on a flat surface, although the
surface may be tilted, such as on a slanted roof,
and may even be mobile, allowing for tracking of the sun throughout the day. Sunlight is
scattered when it hits air molecules, aerosols
and cloud droplets, and is absorbed primarily
by atmospheric water vapor. For flat panels,
whether fixed or tracking, forecasting of the
total amount of radiation reaching the surface
in the relevant wavelength bands becomes critical. Verification of forecasts of radiation reaching a tilted surface are limited by the scarcity of
measurements of solar radiation other than for
radiation reaching a horizontal flat surface. Figure 2a shows solar irradiance traveling through
the atmosphere, being absorbed, scattered,
reflected and some of it landing on the tilted,
flat surface of a PV panel.
and Prediction System (LAPS, laps.noaa.gov).
A number of businesses are using satellite measurements with ground-based measurements
to create solar forecasts (Perez, R. et al., 2007;
Gioioso, M. et al., 2011).
To assess forecast quality, forecasters can also
run multiple weather models, called an ensemble of models. The degree to which the model
outputs in an ensemble differ from each other
indicates the uncertainty in the forecast. Generally, the greater the differences among various
model outputs, the greater the uncertainty in
the forecast. Decision makers, including utilities
and power grid operators, need better information about forecast uncertainty. Therefore, this
“ensemble” technique can be applied to forecasts
of cloud cover, cloud type, aerosols, water vapor
and their impacts on solar irradiance.
More Irradiance Data,
Better Modeling needed
CSP technologies use the direct normal irradiance (DNI) because they focus the direct solar
beam onto concentrating mirrors in order to cre-
Copyright © 2011 by the American Solar Energy Society Inc. All rights reserved.
solartoday.org SOLAR TODAY September/October 2011 55
ate high temperatures. Forecasting of DNI has
been very limited, because there was little need
prior to the development of CSP power plants.
Development of DNI forecasts is hampered by
the paucity of high-quality measurements of
DNI. There are fewer than three dozen publically available, well-maintained, ground-based
measurements of DNI in the United States,
and almost half of these are in Oklahoma at the
Department of Energy’s (DOE’s) Atmospheric
Radiation Measurement (ARM) sites. Though
measurements of horizontal irradiance are more
numerous, for many monitoring sites, the quality is unknown and uncertainties may be quite
large. Figure 2b shows solar irradiance traveling
through the atmosphere, being absorbed, scattered and reflected, with the direct beam landing
on a CSP trough.
The first step to providing an adequate operational solar forecast is to identify and quantify the uncertainties of current solar forecasts
across a range of time scales. Key to improving
solar forecasts is improving our understanding
of how many factors can affect solar radiation.
These factors include aerosols, a large variety of
cloud types and amounts, as well as water vapor,
