Resumen

GOMEZRODRIGUEZ, Marco Antonio et al. Electrical generation forecast of photovoltaic systems. First steps by Cuban Universities. Universidad y Sociedad [online]. 2021, vol.13, n.1, pp.253-265.  Epub 02-Feb-2021. ISSN 2218-3620.

Solar photovoltaic generation is marked by high variability due to the intermittency of solar radiation and other climatic parameters. This makes generation planning difficult. Therefore, accurate short-term forecasting of this type of generators is a crucial factor for power systems. This paper explains the efforts made by Cuban experts and academics in the field to develop predictors within the framework of the Connecting Knowledge project, jointly conducted with Marta Abreu Central University of Las Villas and Carlos Rafael Rodriguez University of Cienfuegos. It describes a hybrid model that combines wavelet transform with artificial neural networks to forecast photovoltaic power generation for the following day, by analyzing recorded data provided by the Data Monitoring and Collection System and local meteorological variables. Several generalized regression and feedforward backpropagation neural networks were developed in Matlab environment; their parameters were altered in order to determine the one that performed the best. The resulting model was developed and validated for a 5.5 MW photovoltaic generation park, which accuracy was compared with the generalized model, revealing improvements rates in the range of 6.66% to 49.71%.

Palabras clave : Photovoltaic generation; forecasting models; artificial neural networks; wavelet transform; renewable energy sources.

        · resumen en Español     · texto en Español     · Español ( pdf )