A data mining system for predicting solar global spectral irradiance. Performance assessment in the spectral response ranges of thin-film photovoltaic modules

Abstract

Knowing the spectral distribution of solar radiation is required to estimate the performance of photovoltaic modules, especially for thin-film modules. This is not a trivial problem due to the large number of environmental factors that affect this distribution as solar radiation passes through the atmosphere. The use of techniques of artificial intelligence and data mining can help in the development of models to address this problem. A system based on these techniques is proposed to predict the solar global spectral irradiance requiring only a few meteorological variables as inputs. The evaluation of the proposed system has been carried out for different wavelengths taking into account the spectral response of different technologies of thin-film photovoltaic modules. The errors in predicting solar global spectral irradiance for wavelengths that range between 350 and 900 nm and air mass lower than 2.1 are smaller than 7% on clear-sky days and than 16% for cloudy days, which is a significant improvement on other proposed models. Moreover, an open access implementation of the developed system is available at the URI: http://fvred1.ctima.uma.es. It could be useful for engineers and companies in the fields of the environment and renewable energies.