Incoming solar radiation, or irradiance, is a key attribute in the research for sustainable energy resources. It is typically measured indirectly, and the observations are known to contain significant uncertainty. To study its spatiotemporal distribution, classical geostatistical methodologies are often used that are unable to account for that uncertainty.
Our analysis proposes meaningful assimilation of observed data by using the measurement uncertainty as part of the information with the Bayesian Maximum Entropy theory. As an example, we predict distributions of monthly-averaged daily global solar irradiance on a continental scale across the USA for every month in 1999. A similar approach is applicable in more localized studies for photovoltaic system prediction, where space-time prediction is desired for smaller spatial grids, and over finer temporal resolutions in the order of hours or minutes.
In the following animated maps we illustrate the predicted monthly-averaged daily global solar irradiance (mdGSI, measured in MJ per square meter per day) across the USA for every month in 1999. The values of mdGSI range from 0 (white) to a maximum of 30 (dark brown).
This is a proof-of-concept study on a coarse continental grid, for which we used information only from monitoring stations across the country that recorded direct, non-modeled solar radiation values. The original hourly global solar irradiance (GSI) observations have a documented level of about 6% uncertainty. We aggregated this information to monthly averages and their corresponding uncertainties for every month in 1999. Consequently, these soft probabilistic data were the input for the spatiotemporal prediction of mdGSI.
Irradiance observations are most commonly derived from secondary meteorological observations, radiation models or satellite-measured quantities. The expected error in these estimates can reach a working level of 10% for satellite estimates. Our Knowledge Synthesis framework techniques enable rigorous account of the observation uncertainty in space-time predictive process to yield more accurate results. You can use this approach to improve accuracy and increase reliability in your photovoltaic predictive models, feasibility studies, investment plans, energy pricing policies.
Kolovos A. 2009. Spatiotemporal Analysis of Solar Radiation for Sustainable Research in the Presence of Uncertain Measurements. Proceedings of the StatGIS 2009 Conference. Milos, Greece, June 2009.