Dax Matthews, Larry Cutshaw and Richard Rocheleau
Hawaii Natural Energy Institute, United States
Keywords: solar, forecasting, photovoltaic,satellite, skyimager
The integration of photovoltaics into the grid poses problems due to natural variability of the resource. Clouds are the primary factor in modulating insolation by attenuating solar beam irradiance. A forecasting system that predicts cloud locations and solar irradiance at minute to hour time scales, employing a unique combination of ground-based skyimagery and Geostationary Operational Environmental Satellites (GOES) radiometry is presented. Clouds are identified in skyimagery using the red-blue-ratio threshold method and in GOES imagery using the bispectral composite threshold technique. Three-dimensional cloud maps are generated from a coordinate transform of the ground and satellite imagery, utilizing cloud height information from ceilometer and GOES sounder data. Cloud velocities, determined from the maximum cross-correlation technique, are used to forecast cloud locations by horizontal advection of the cloud maps. Cloud shadows are determined from the cloud maps and surface irradiance is estimated by a clear-sky radiative transfer model. Validation of the forecasting system using pyranometer ground stations on the islands of Oahu, Maui, and Hawaii is also presented. Semivariogram and swinging door algorithms are used to determine the amplitude, frequency, persistence, and time scales of fluctuations in pyranometer irradiance time series, which are used to tune and evaluate the forecasting system.