Evaluation of some linear regression models for the prediction for daily global solar radiation in the village of Koyli Alpha, district of Linguere, Louga region, Senegal

In this paper, we compare three models in order to determinate the daily mean monthly of global solar radiation using climate conditions of a Koyli Alpha, which is a village located in the district of Ferlo zone, North of Senegal. The first model used the temperature, the model 2 depends on the sunshine duration and the model 3 is related on the sunshine duration and the relative humidity. These models are calculated using both the regression coefficients and the meteorological data recorded from the set-up weather station in Widou village. This village is located about 20 km from the village of Koyli Alpha. Both villages have neighboring geographical characteristics and are not far apart. The best model for the estimation of the monthly mean daily global solar radiation on horizontal surface in Koyli Alpha is determined. Hence, it was observed that the highest predicted values, measured in April, year 2017, are 5,456.26 Wh/m²/day for the model 1, 5,521.80 Wh/m²/day for model 2 and 5,450.67 Wh/m²/day for the model 3. The first model is better than models 2 and 3 because it has the highest values of the coefficient of determination (R²), NSE, SDFE and the lowest values of MRE and RMSE. Among these three linear regression models, the first model using ambient temperature is the most accurate model for the site of Koyli Alpha because it is justified by the statistical parameters used in this paper.