Photovoltaic energy prediction using multi-layer neural networks

摘要

The power generated by photovoltaic (PV) systems depends on meteorological parameters, such as the irradiance and the outdoor temperature, along with the presence of clouds. Moreover, it depends on some cell parameters, such as the cell temperature. Due to the continuous change in power, it can be important to predict the energy production of PV systems, starting from some basic input parameters. The aim of this paper is to show that a multi-layer perceptron neural network can represent a useful tool to carefully predict the PV energy output. By using two different sets of experimental data (collected every thirty minutes during some days in June 2009 for a mono-crystalline silicon module) and by exploiting a back propagation learning algorithm, two suitable neural architectures are found, one for each data set. For both cases simulation results show that the estimated PV energy values (predicted by the proposed network) are in good agreement with the experimental measured values, thus providing a valuable tool for PV end-user.