Performance Evaluation and Comparison of Models for Predicting Global Solar Radiation in Ethiopia: A Case Study of Two sites

Accurate estimation of the global solar radiation on a horizontal surface at a place is essential to design and check the performance of solar energy devices. This study focused on evaluating monthly average daily global solar radiation (DGSR) and monthly average global solar radiation (GSR), as well as comparing the six models used in the estimations, over Lalibela and Sirinka in Northern Ethiopia. The best models were selected according to the values of statistical performance evaluation. For Lalibela the best model to estimate DGSR was LO-model with values of MBE=-0.092 KWh/m2day, RMSE= 0.402 KWh/m2day, MPE=-1,969 % and for Sirinka- GM model with values of MBE=-0.575 KWh/m2day, RMSE= 0.643 KWh/m2day, MPE=-9.764 %. The best estimator of monthly average GSR over Lalibela was AP model with values of MBE=0.159 KWh/m2day, RMSE= 0.245 KWh/m2day, MPE=2.802 % and for Sirinka- LO model with values of MBE=- 0.007 KWh/m2day, RMSE= 0.522 KWh/m2day, MPE=0.033 %. At Lallibela site, the estimated monthly average DGSR range from 4.38 KWh/m2 (in August) to 7.16 KWh/m2 (in March) and the monthly average GSR estimated is 4.96 KWh/m2 (in July) and 6.69 KWh/m2 (in March). The estimated monthly mean DGSR range from a 3.80 KW/m2 (in January) to a 5.97 KW/m2 (in May) while the predicted monthly mean GSR varies from a 4.48 KW/m2 (in January) to a 6.75 KW/ m2 (in May) for Sirinka site. Most importantly, results show that estimation of DGSR and GSR not only takes into account meteorological data but also includes the latitudinal difference between the sites (GM model) and values of Nnh (LO model). The overall results suggest that the sunshine based empirical models had better performance compared with the temperature based empirical models for both study sites. Finally, Lalibela site has a higher global solar radiation potential than Sirinka.