Artificial Neural Network Model for Precise Estimation of Global Solar Radiation

A new model based on artificial neural network has been developed to estimate monthly average global solar radiation for different Indian locations. Two layers feed forward network sigmoid trained with Levenberg-Marquardt back propagation algorithm with eleven input terminals and ten hidden layer neurons have been used to give solar radiation as an output. The model was trained, validated and tested by using measured global solar radiation data of eighteen Indian locations spread over different Indian climatic zones for which measured data was available. The regression coefficient was found to be 0.95967 with mean square error of 0.204. The mean percentage error, root mean square error and mean bias error between estimated and measured global solar radiation of eighteen locations have been found to be in the range of -4.16 to 4.82 , 0.02 to 0.26 and -0.30 to 0.08 respectively. A graphical user interface has also been developed to find the monthly global solar radiation of any location throughout India by putting eleven input geographical parameters of the desired location. Research Highlights: • Development of an artificial neural network model for the estimation of monthly average global solar radiation • Graphical user interface development for the developed artificial neural network model • Graphical representation of best validation performance • Comparison of measured and estimated solar radiation data • Regression coefficient representation