Photovoltaic generation prediction using the deep learning long short term memory model

Abstract

Solar photovoltaic energy is a renewable, clean and safe energy source that is currently used worldwide. However, it presents a dynamic and intermittent behavior, caused by the variation of climatic conditions. Due to this, it has been necessary to develop different methods for the prediction of the energy generated in photovoltaic systems. The present work is focused on analyzing the prediction of the power generated in a photovoltaic plant connected to the grid, by means of the Long Short-Term Memory (LSTM) deep learning model. In order to carry out the study, a database obtained from the photovoltaic plant of the Central University "Marta Abreu" of Las Villas (UCLV) with a nominal installed power of 1.1 MW is used. Initially, the correlation between the different variables with respect to the photovoltaic power generated is analyzed, then the LSTM model is implemented to make the prediction. The results obtained show that the predictions made for different time horizons and for days with different behavior are adequate, which demonstrates the effectiveness of this prediction method.