Empirical Validation of Regression Models for forecasting Global Solar Irradiance: A Case Study of Abeokuta, Nigeria

Abstract

The performance of solar radiation distribution models is highly dependent on the location of use. This makes the need for location dependent validation necessary. An empirical validation of three solar radiation distribution models was carried out in this work using empirical data obtained from the Nigeria Meteorological Agency (NIMET). The models include the Angstrom-Prescott, Hargreaves and Garcia model. Four standard statistical tools of correlation ratio (R), mean bias error (MBE), mean percentage error (MPE), root mean square error (RMSE) were used for the validation. Results from the work reveal that the Angstrom-Prescott model had the least MBE with an average value of 0.1 making it the most reliable among the three models. In terms of MPE values, the Angstrom-Prescott model also performed best with an average MPE value of 0.1 while the Hargreaves and Garcia model had an average MPE performance of 0.2 and 0.15 respectively. In terms of the RMSE and correlation ratio, the Angstrom-Prescott model also performed better than the other two models with an average RMSE and correlation ratio of 0.3 and 0.65. This work thus reveals that, for a medium term solar radiation forecasting goal, the best regression based model that accurately predict the solar radiation distribution in Abeokuta is the Angstrom-Prescott model.