Prediction of cloudiness in short time periods using techniques of remote sensing and image processing

Abstract

In this work we introduce a methodology which enables to predict the cloudiness in the short and medium termanywhere in the world. Satellite images (Meteosat of Second Generation) are used in combination with images from a sky camera (fisheye lens), showing a ground vision of the clouds, and using the real-time radiation measured on-site as a feedback and as a complement to the cloudiness. The methodology is based on the determination of cloud motion in the images. Obtaining cloud movement vectors from consecutive images, we are able to anticipate the displacement of the previously detected clouds, thus knowing the distribution of clouds in the future. The short-term forecast (less than 1 hour) and the medium-term forecast (up till 3 hours) have a rate of success of 80%. Aiming to have an accurate knowledge of the evolution of cloudiness in the short and medium-term (useful for CSP plant management) an interactive portal has been developed. The application is a user-friendly interface which shows three hours real-time forecasts refreshed each minute, along with useful information for the operation of the CSP plant like the DNI evolution, the original and processed image from satellite MSG-2 as well as that from sky camera. In the application 400 Wm-2 will be considered as the DNI threshold for the optimal operation for a CSP plant. The application has been tested and validated in two different locations: University of Almería (Almería, Spain) and Gemasolar Central Tower Plant (Fuentes de Andalucía, Spain) and it is going to be installed in Valle 1 and 2 Parabolic Trough Plant (San José del Valle, Spain).