Improvements of short-term solar radiation forecasting techniques based on satellite and sky-camera imagery

Abstract

The energy resources dependant of weather generate energy production varibility, which difficult its integration into the electricity grid. The prediction of resources aims at anticipating contingency actions in order to manage this variability. In this thesis, improvements in nowcasting and short-term solar forecasting methods are studied, particularly, those based on sky-camera and satellite imagery. This thesis includes: 1) the detailed description and development of state-of-the-art solar radiation forecasting methods, 2) the use of sky-camera as automatic cloud type classificator, 3) an analysis of weather patterns associated to several modes of solar variability, 4) a comparative analysis of several short-term solar radiation forecasting methods, and 5) the proposal and assessment of blending models computed through machine learning techniques.