Estimating daily diffuse solar radiation using neuro-fuzzy, fuzzy C-means and subtractive clustering - Case study - Adrar, Algeria.

This work presents two optimization techniques, fuzzy c-means (FCM) and subtractive clustering, to improve the efficiency and generalization ability of back propagation neuro-fuzzy model for estimating daily diffuse solar radiation Hd, four parameters including extraterrestrial irradiation H0, the global radiation Hg, the average, air temperature (Tavg) and the relative humidity (Rh). The predictions from the neuro-fuzzy optimized by fuzzy c-means (FCM) and subtractive clustering were analyzed and evaluate with 02 statistical parameters. The results show that the proposed neurofuzzy optimized by FCM model has potential in accurately predicting the daily diffuse solar radiation.