Prospective energy generation in photovoltaic systems, based on the comparison and evaluation of different configurations based on artificial neural networks (ANN)

Ramos Cutipa J.M.; Condori Yucra R.; Shuta Llocclla H.; Chayña Velasquez O.; Paredes Pareja W.O.; Vilca Choque M.C.; Beltrán N. · Proceedings of the LACCEI international Multi-conference for Engineering, Education and Technology (2022) · DOI: 10.18687/laccei2022.1.1.719

Resumen

For photovoltaic systems to be sustainable in the time, it is necessary to have a better use of solar energy, so probabilistic predictions of variables that directly influence the production of electrical energy must be taken into account when making decisions. Solar irradiation is the important variable for these predictions, which allows managing the production of photovoltaic energy. In the present work, the application of RNA artificial neural networks properly configured to interact with the information that is available on climatological variables is proposed in order to be able to predict solar irradiation. The correlation that exists between these variables have been analyzed in order to be able to compare different configurations of MLP-type neural networks for one variable and multiple variables, in addition to using GRU-type recurrent neural networks. According to the MAE, the GRU type networks are the ones that provide the best results, however, considering the ease of training, the MLP type networks are acceptable, being the multivariable type of the ones that give the best results. Looking to the future, the use of these networks using pattern recognition in sky images considering the movement of the clouds and the earth with respect to the sun is proposed.

Instituciones: Universidad Nacional del Altiplano