Intra-hour forecasting with a total sky imager at the UC San Diego solar energy testbed

Accurate solar forecasting has become increasingly important as the solar energy supply is expected to grow. The ability to predict the fluctuating nature of the solar resource affects the economic viability of solar energy generation. Currently, short-term solar forecasting is performed primarily using satellite-based techniques through cloud tracking. However, the spatial and temporal resolution of geostationary satellite images is not sufficient to provide information on intra-hour solar variability, which has a significant impact on operation of solar thermal power plants, energy storage, and load balancing on the electricity grid. Improved monitoring of the solar resource on smaller spatial and temporal scales is required to provide this intra-hour information. A ground-based total sky imager can provide continuous monitoring of the sky hemisphere above a site. In this paper, we present a method to forecast global horizontal irradiance using fractional sky cover and cloud motion vectors. Sky images taken every 30 seconds are processed to determine sky cover. Cloud motion vectors are generated by cross-correlating two consecutive sky images. Future cloud locations are then computed by applying the vector field to the sky scene. Using the updated cloud locations global horizontal irradiance at the surface is computed and validated using a dense network of eight pyranometer ground stations measuring every second.