The improvement and comparison of diffuse radiation models in different climatic zones of China.

Diffuse radiation is vital for climatology, sustainable energy, agriculture and biological activities. It is often estimated based on some meteorological factors but seldom takes total cloud cover into consideration. In this study, cloud cover data are used to establish models, compared with other regular models for estimating diffuse radiation in different climatic zones of China during 1992–2015. The results showed that the performance of the new model based on sunshine fraction (n/N), clearness index (k t), total cloud cover (CI), air temperature (T a), relative humidity (RH), wind speed and the day of the year outperformed other models and further improved the predictive accuracy. Generally, the k d models based on the single k t was generally more precise compared with single n/N-based or single CI-based models. And models on the basis of multiple factors performed better than the single factor-based model. Comparisons between these models indicated that neural network models (with the largest R and smallest RMSD values) provided better overall accuracy than other models. The parameter of CI as the co-factor can improve the prediction of diffuse radiation. The models developed and evaluated in this study can contribute to developing and utilizing solar energy in China, especially in areas without diffuse radiation records. [Display omitted] • The work can estimate diffuse radiation accurately, especially in some areas without diffuse radiation measurement. • Applicability of 21 empirical and 2 artificial neural network models were investigated in five climatic zones of China. • The models considering diffuse fraction, sunshine fraction and cloud cover and four parameters were proposed. • The parameter of total cover cloud as the co-factor can improve the prediction of diffuse radiation. [ABSTRACT FROM AUTHOR]