A neural network model for estimating global solar radiation on horizontal surface

Abstract

This research focuses on the development of artificial neural network (ANN) model for estimation of daily global solar radiation on horizontal surface in Dhaka. In this analysis back-propagation algorithm is applied. Day of the year, daily mean air temperature, relative humidity and sunshine duration were used as input data, while the daily global solar radiation was the only output of the ANN. The database consists of 1827 daily measured data, between 2008 and 2012, in term of daily mean air temperature, relative humidity and sunshine duration and global solar radiation. The data has been collected from Bangladesh Meteorological Department. The 1461 daily measured data between 2008 and 2011 are used to train the neural networks while the data of 366 (leap year) days from 2012 are used to test the neural network. MATLAB neural network toolbox is used to train and test the network. Both estimated and measured values of daily global solar radiation on horizontal surface were compared during testing phase statistically using two methods: Root Mean Square Error (RMSE) and Regression R Value (R), giving a value of 113.6 Wh/m2 and 0.9744, respectively. The results of this study have shown a better accuracy than other conventional prediction models that have been used up to now in Bangladesh. This ANN model may be suitable for predicting solar radiation at any location in Bangladesh, provided that samples of the sunshine duration data from the locations are available.