A REVIEW OF DAILY GLOBAL SOLAR RADIATION MODELLING USING DIFFERENT STATISTICAL METHODS BASED ON SUNSHINE DURATION IN GRAN CANARIA ISLAND

摘要

Recently, solar photovoltaic and thermo electric systems have become one of the most important renewable energies. Global solar irradiation (GSR) is a very important variable for developing solar energy production in a region. GSR description and modeling requires solar radiation measurements in a wide range of location around this region. The high costs of solar radiation measurement stations avoid its installation in big areas around the world. Since years, many different GSR estimation models have been developed using several meteorological variables. Many papers can be found using sunshine duration, temperature, relative humidity and air pressure, which in fact are variables easier measured. In this paper, different Angstrom models, based on sunshine duration, have been tested using measurement stations around Gran Canaria Island following previous works. The Island presents very different climatological behavior depending on the location because of its complicated orography and the influence of Trade Winds. In order to compare classical Angstrom models results, Artificial Neural Networks are proposed for modeling daily GSR using the day of the year and the sunshine duration as inputs. It is observed that for all measurement stations, neural networks outperform classical models using different network architectures.