FPGA-based artificial neural network for prediction of solar radiation data from sunshine duration and air temperature

Abstract

The software as well as the hardware implementation of solar radiation prediction based on Artificial Neural Network (ANN) and Hardware Description Languages (HDLs) are presented in this paper. It introduces preliminary results of the Filed Programming Gate Array (FPGA) implementation of solar model. FPGA technology was employed due to its development flexibility and low cost. A database comprising a set of meteorological data such as: solar radiation, temperature and sunshine duration have been used in this study. The inputs of the ANN-solar model are the sunshine duration and mean average temperature while the output is the daily total solar radiation. Firstly, a dataset of several sites have been used for training the network. Subsequently, after the design procedure was done, the developed architecture by VHDL is then implemented on reconfigurable FPGA device (Xilinx, VirtexII). The developed hardware device permits the forecasting of the daily solar radiation using available mean temperature and sunshine duration, especially in isolated sites, where there are no instruments for measuring the solar radiation data.