Sistema para predicción de la generación en bloques de plantas fotovoltaicas.

The prediction of energy generation in photovoltaic plants connected to an electrical system is the subject of constant study and development. The objective of this work is to develop a method for short-term generation prediction. This prediction is made from the application of prediction programs for photovoltaic plants developed based on recurrent and convolutional neural networks. Solar irradiation and ambient temperature are considered as input data. Experimental methods are applied, using historical data on the behavior of these variables, and applying different methods of pre-processing these data and post-processing the basic predictions, which provides predictions with greater accuracy. The main result is a prediction method, with better accuracy in the central hours of the day. The results of the predictions for a group of plants are presented, demonstrating the feasibility of applying the method. [ABSTRACT FROM AUTHOR]