Calculation of the energy provided by a PV generator. Comparative study: Conventional methods vs. artificial neural networks

The use of photovoltaics for electricity generation purposes has recorded one of the largest increases in the field of renewable energies. The energy production of a grid-connected PV system depends on various factors. In a wide sense, it is considered that the annual energy provided by a generator is directly proportional to the annual radiation incident on the plane of the generator and to the installed nominal power. However, a range of factors is influencing the expected outcome by reducing the generation of energy.The aim of this study is to compare the results of four different methods for estimating the annual energy produced by a PV generator: three of them are classical methods and the fourth one is based on an artificial neural network developed by the R&D Group for Solar and Automatic Energy at the University of Jaen.The results obtained shown that the method based on an artificial neural network provides better results than the alternative classical methods in study, mainly due to the fact that this method takes also into account some second order effects, such as low irradiance, angular and spectral effects.

Research highlights► It is considered that the annual energy provided by a PV generator is directly proportional to the annual radiation incident on the plane of the generator and to the installed nominal power. ► A range of factors are influencing the expected outcome by reducing the generation of energy (mismatch losses, dirt and dust, Ohmic losses,.). ► The aim of this study is to compare the results of four different methods for estimating the annual energy produced by a PV generator: three of them are classical methods and the fourth one is based on an artificial neural network. ► The results obtained shown that the method based on an artificial neural network provides better results than the alternative classical methods in study. While classical methods have only taken into account temperature losses, the method based in an ANN has taken into account temperature losses, low irradiation losses and spectral and angular losses.