Assessment of characteristics of surface solar irradiance on consecutive days using a self‐organizing map and clustering methods

Abstract

Abstract The assessment of surface solar irradiance as a resource is a major research topic in renewable energy. A method for assessing the characteristics of global horizontal solar irradiance (GHI) at the surface on consecutive days is proposed in this work. A self‐organizing map and cluster analysis are used as core analysis techniques in the proposed method. Using this method, the regional and seasonal characteristics of GHI for five consecutive days are analysed. One hour accumulated GHI data from ground observation stations in Japan are used. Time series assigned to each node on the topological map of the self‐organizing map are characterized by strength and variability. The probabilities of occurrence of each node on the map differ between stations. Then, using the probabilities as a feature vector for quantifying the characteristics of each observation station, a cluster analysis is conducted to divide stations into clusters according to the dissimilarity of the features. The results show that there are three major regions in which the station characteristics are similar in Japan. Additionally, another cluster analysis is conducted to investigate the seasonal characteristics of the occurrence of time series patterns. It is found that consecutive cloudy days occur frequently in winter and during the rainy season, whereas consecutive clear days occur frequently in spring and summer.

Keywords

• cluster analysis
• discrimination
• energy
• self‐organizing map
• surface solar irradiance
• transport