Novel models to estimate hourly diffuse radiation fraction for global radiation based on weather type classification. (September 2020)

• Record Type: Journal Article
• Authors: Li, Fen; Lin, Yilun; Guo, Jianping; Wang, Yue; Mao, Ling; Cui, Yang; Bai, Yongqing
• Is Part Of: Renewable energy. Volume 157(2020)
• Page Start: 1222
• Page End: 1232
• Publication Date: 2020-09
• Subjects: Diffuse radiation fraction -- Weather type classification -- Modified clearness index -- Principal component analysis -- Combined forecast model -- Typical meteorological year; Renewable energy sources -- Periodicals; Power resources -- Periodicals; Énergies renouvelables -- Périodiques; Ressources énergétiques -- Périodiques; 333.794
• DOI: 10.1016/j.renene.2020.05.080
• Languages: English
• ISSNs: 0960-1481

Abstract

Abstract: The diffuse radiation is well recognized as a key variable in solar energy assessment, albeit with sorely lacking ground-based measurements. Here, we proposed two novel models to estimate hourly diffuse radiation using the typical meteorological year's radiation data in Beijing as training samples. Model 1 was a combination of four classical models, including Liu&Jordan, Orgill&Hollands, Erbs and Reindl, in which the weight or coefficient was determined by weather types that were derived from clearness index. In Model 2, the weather type classification was refined by total cloud cover, and the principal component analysis (PCA) was further applied to determine the major meteorological variables for each weather type as model's input, along with linear fitting. Using sub-typical year's radiation data as testing samples, the proposed models showed strong extrapolation ability with three statistical metrics: lower mean absolute percentage error and normalized root mean square error but relatively higher correlation coefficient, compared with other models. Finally, these models were verified by the observations in Wuhan. The results indicated that weather type classification and PCA effectively improved model's performance by eliminating the collinearity between meteorological and environmental variables. Furthermore, both models performed better than any single classical model, irrespective of large-scale weather patterns. Highlights: Novel diffuse radiation model proposed using weather type classification. Typical meteorological year is selected based on long-term measurements in Beijing. Principal component analysis (PCA) used to select major variables as model inputs. Weather type classification and PCA can effectively improve model's performance.