Research on solar beam-diffuse radiation separated model based on BP neural network

Trying to PM
₂
.s as the independent variable to establish the “beam-diffuse radiation separated” model, factor analysis showed that, PM
₂
.
₅
and diffuse ratio have positive correlation. Based on the clearness index K
_t
and sunshine hours n/N as variables, the polynomial model and the BP neural network algorithm model containing PM
₂
.
₅
as the independent variable is proposed, the polynomial model is fitted and the BP neural network model is trained using the 5 years measured data. Using the established model to predict the diffuse ratio H
_d
/H in June 2013, combined with the measured data, error analysis using statistical indicators, the results show: the BP neural network model based on clearness index K, sunshine hours n/N and PM
₂
.
₅
prediction accuracy is highest.