Solar energy forecasting using a total sky imager

A major problem for solar energy is the occlusion of the sun by clouds. It introduces variability in power output that is difficult to mitigate. To solve this problem, short-term forecasts of the motion of clouds through the sky can be made. For a sequence of sky images, the motion detected between two images is used to warp the second image to an estimate of the third image, and forth, and so on. Many methods exist for estimating motion in sky images. This work endeavors to develop approaches to outperform those methods using a number of previously untested techniques, such as inpainting, optical flow, blob detection, and clear-sky polynomial forecast models.