Clustering daily solar radiation from Reunion Island using data analysis methods

Abstract

A partitioning of daily solar radiation using Hierarchical clustering on Principal Components method is proposed based on daily distribution of direct fraction in solar irradiance. This tool, similar to clearness index, is only based on diffuse and global solar radiation measurement, taking in account the local topology. The clustering method is applied on data collected in Reunion island (21◦S, 55◦E), a southern subtropical site. We obtain five daily solar radiations classes as results. The interpretations of the average time series signals for each class shows that daily dynamics curves are correlated with local meteorological phenomena. The season distribution of classes is correlated to the intensity of trade winds f low. Thus, it helps us to define an approach to predict the solar radiation for different time horizons and the energy potential of the day.