Evaluation of sunshine-based models for predicting diffuse solar radiation in China

Accurate observation and understanding of diffuse radiation is of vital importance for solar energy applications. Numerous empirical models have been developed for estimating solar radiation in regional and global scales, owing to the relatively sparse radiation measurements. The main objective of this study was to conduct a comprehensive evaluation of 15 typical empirical models for estimating diffuse radiation in different climate zones over mainland China. The result showed that the model in form of second order polynomial performed superior than other models, with mean MBE, MAE, MARE, RMSE, RRMSE, t-stat, STD, and R at all 17 CMA stations were -0.125 MJ m(-2) day(-1), 1.331 MJ m(-2) day-1, 0.208 MJ m(-2) day(-1),1.807 MJ m(-2) day(-1), 24.889%, 10.866, 0.941 MJ m(-2) day-1, and 0.792, respectively. By contrast, the model in form of fractional first order polynomial showed the poorest performance than other models, with mean MAE, MARE, RMSE, RRMSE, t-stat, STD, and R of-0.699 MJ m(-2) day(-1), 2.508 MJ m(-2)day(-1), 0.397 MJ m(-2) day(-1), 6.779 MJ m(-2) day(-1), 102.716%, 6.709, 1.773 MJ m(-2) day(-1), and 0.519, respectively. All models generally showed poor accuracies in arid areas with warm-temperate climate, due to the frequent dust occurrences in the air. The estimation errors in Qinghai-Tibet Plateau were also relatively larger, owing to the strong heating atmosphere there. This study would assist in the selection of the most appropriate models for solar energy applications.