Long-Term Global Solar Radiation Prediction in 25 Cities in Morocco Using the FFNN-BP Method

This article presents different combinations of input parameters based on an intelligent technique, using neural networks to predict daily global solar radiation (GSR) for twenty-five Moroccan cities. The collected measured data are available 365 days and 25 stations around Morocco. Different used, such as clearness index K T , day number, the length day, minimal temperature min maximal max average difference ΔT ratio T-Ratio, relative humidity RH, at top outside atmosphere TOA, wind speed Ws, altitude, longitude, latitude, declination. A combination was employed GSR considered locations in order find most adequate parameter that can be used prediction procedure. Several statistical metrics applied evaluate performance obtained results, coefficients determination ( R 2 ), mean absolute percentage error (MAPE), root square (RMSE), normalized (NRMSE), bias (MBE), test statistic (TS), linear regression (the slope “a” constant “b”), standard deviation (σ). It is found usage gives highly accurate results artificial network (FFNN-BP) model, obtaining lowest value metrics. showed best locations, 12 inputs Er-Rachidia, Marrakech, Medilt, Taza, Oujda, Nador, Tetouan, Tanger, Al-Auin, Dakhla, Settat, Safi, seven Fes, Ifrane, Beni-Mellal, Meknes, six Agadir Rabat, five Sidi Ifni, Essaouira, Casablanca Kenitra, four Ouarzazate, Lareche, Al-Hoceima. In terms accuracy, selected varies between 0.9860% 0.9920%, range MBE (%) being from −0.1076% −0.5931%, RMSE 0.1990 0.4580%, NRMSE 0.0355 0.8938, MAPE 0.0019 0.0060%. technique could other where measurement instrumentation unavailable or costly obtain.