Prototyping and Validating a Solar Irradiance Forecasting System

As solar and wind energy resources supply a greater share of electricity production, it will become increasingly necessary to develop controls to handle these unstable energy sources and balance supply cycles that are at times out of sync with demand cycles. These power grid controls must enable maximum exploitation of the renewable energy sources when they are active and productive, while allowing traditional power plants to provide energy during off-peak production times, for example overcast skies and light winds. Detection (nowcast) and forecast of these changes will provide information which allows power production from conventional power plants to be adjusted to meet demand.

We have developed a prototype system for solar irradiance forecasting which integrates cloud remote sensing, mesoscale forecast modeling and radiative transfer technologies. The resulting product can forecast solar irradiance on hours-to-days time scales for larger-scale weather patterns, and minutes-to-hours time scales for fine load balancing adjustments between solar and other power generation sources. The largest source of ground-level solar irradiance variability is due to variations in cloud cover. This paper will present the details of our solar forecasting system, with emphasis on the accurate prediction of cloud cover and the associated ground-level solar direct and diffuse components, and discuss benchmark statistics.