Estimación de la potencia producida por módulos fotovoltaicos en base a Redes Neuronales Artificiales y comparación con metodologías basadas en Regresiones Múltiples

Abstract

Solar irradiation has recently acquired huge importance as the main source of energy for the present and future. Therefore, several analyses have to be carried out in order to determine the actual generation expectancy of photovoltaic plants. The forecasting of the potential output power that can be obtained from a photovoltaic module then derives in a great amount of applications of engineering, as planning and installation of photovoltaic plants. In this connection, application of predictive algorithms that give an approximation of expected generation, presents a challenging issue for power engineers. Thus, Artificial Neural Networks (ANN) can be used for modeling the behavior of the photovoltaic generator, through the data mining of his most influent variables: solar irradiation and module temperature. This paper shows the procedure for performing a robust training in the most accurate Artificial Neural Network in order to accomplish an accurate prediction of the expected generation of CENACE’s photovoltaic plant. The obtained results, based on ANN model, will be then compared with those obtained using methods based on multivariate regressions.