Assessment of Five Predictive Models for Solar Radiation in Southwest Nigeria

  • Alexander Willoughby Redeemer's University
  • Amanda Ndubuisi Covenant University
  • Oluropo Dairo Redeemer's University
  • Ayodele Soge Redeemer's University
Keywords: Annual variation, meteorological variables, energy, performance indicator, statistical test


This study compares the accuracies of five predictive models for estimating solar radiation amongst other meteorological parameters in Southwest Nigeria. Twenty-one years of monthly averages of six measured meteorological parameters obtained from six stations in southwest Nigeria have been subjected to five mathematical models for prediction purposes. Solar radiation and sunshine hours have been modelled using the sum of two-Gaussians, the sum of two-Lorentzians, Fourier on four harmonics, sine wave and fourth-order polynomial functions. The fitting accuracies of these models were tested using performance indicators; mean bias error (MBE), root mean square error (RMSE), mean percentage error (MPE), standard error (SE) and the correlation coefficient (R). An evaluation of the models showed that the Gaussian and Lorentzian models are in good agreement with the observed data. However, the Fourier on the fourth harmonics model had the lowest MBE, RMSE and MPE, consequently highest correlation coefficient values, indicating high model accuracy. Thus, the Fourier model has the best correlation with the observed data and is recommended for estimating these variables in the selected locations.