CloudGaze: landsat sky image based determination of cloud motion to predict solar obscuration

Abstract

The detection and prediction of cloud motion from ground based visible light images are critical for optimizing solar energy generation and improving solar facility yield predictions. Moreover, understanding cloud dynamics provides valuable insights for climate modeling. This paper presents CloudGaze, a novel system designed to detect and predict cloud movement on to the sun using Landsat images of the sky. The system utilizes image segmentation of the cloud regions using HSV color-space, advanced image processing techniques to extract the largest cloud contour, and the Lucas-Kanade optical flow algorithm to predict the cloud movement, By leveraging these components, CloudGaze aims to provide prior notifications of impending cloud cover on the sun, enhancing solar generation efficiency,…