Machine Learning techniques for solar irradiation nowcasting: Cloud type classification forecast through satellite data and imagery

One of the most important modern challenges in making the renewable energy sources more reliable is the development of new tools to better manage their non programmable nature and avoid economic losses, to ensure compliance with network constraints and to improve the management of congestion. The solar energy at ground level exhibits a continuous variation in time and space. This fluctuation has a deterministic component generated by the movements of rotation and revolution of the earth, and a random one generated by weather conditions. Solar energy variations at ground level have a great influence on the output power of a photovoltaic plant, which can fluctuate significantly in short intervals due to the random component. This work presents a new model to detect in real time the clouds which potentially obstruct the sunrays directed to a specific geographic target. Moreover, a novel procedure for the forecasting of the clearness sky index on the target in the fifteen minutes is proposed, levereging on Machine Learning techniques, exploiting satellite and weather data.