Correlation analysis of solar radiation and cloud cover

Measured hourly solar radiation and cloud cover data were analysed. Regression analysis techniques were used to correlate the monthly-average-daily and monthly-average-hourly global and diffuse radiation with the corresponding cloud cover data. The mean bias error (MBE) is very small, representing less than 1% of the mean measured values. This indicates the regression models developed do not exhibit any long-term persistence in under- or over-estimation of solar radiation from cloud cover data. However, on an individual year basis, the MBE can be as high as 20% of the mean measured values for the hourly data prediction. It has also been found that using multiple regression techniques to correlate solar radiation with both the cloud cover data and the solar altitude can improve the prediction by about 5% as compared with the simple Ångstrom-Prescott linear regression equation. This paper describes the analysis and presents the findings.