An artificial neural network estimation of global solar radiation at Ibadan, Nigeria using meteorological data

Abstract

This paper estimates global solar radiation (Rs) from routinely measured meteorological parameters in the city of Ibadan, Nigeria, using artificial neural network method. Six combinations were used to estimate Rs namely (i) daily mean air temperature (T) and day of the year as inputs and global solar radiation as output, ((ii) daily mean relative humidity (RH) and day of the year as inputs and Rs as output (iii) daily mean T, daily mean RH and day of the year as inputs and Rs as output (iv) daily mean minimum relative humidity (RHmin) and day of the year as inputs and Rs as output, (v) daily mean minimum temperature (Tmin), daily RHmin and day of the year as inputs and Rs as output (vi) daily mean maximum temperature (Tmax), daily mean Tmin, daily mean RHmin, daily maximum relative humidity (RHmax) and day of the year as inputs and Rs as output. The neural network was trained with 3653 measured data between 1995 and 2004 and tested with data for 731 days between 2003 and 2004. The data for testing the neural network were not used for the training. The results obtained showed that the combination of RHmin, RHmax and day of the year gave the best estimate of Rs with MSE of 3.4124. This is followed by RHmin and day of the year with MSE of 3.4424. Daily mean air temperature and day of the year could not mimic the measured Rs; it gave MSE of 5.3345. It is concluded that Rs can be estimated for locations where only temperature and relative humidity data are available.