A Wavelet Neural Network Model for Hourly Solar Radiation Forecasting from Daily Solar Radiation

Abstract

This paper represents a framework for forecasting solar radiation data per hour using mean of solar radiation on daily basis. The proposed model which is used for forecasting is wavelet neural network. Sunset hour angle, hour angle and daily solar radiation are the three inputs of this model. The output layer has one node which is mean hourly solar radiation (HSR). The proposed model has been trained and developed using MATLAB software. The outcomes achieved here have better prediction precision in comparison to some previous empirical models. To evaluate the proposed model, two error statistical method such as root mean square error and mean absolute percentage error has been implemented. Finally R2 estimation of 96% of the proposed model demonstrates that it can be implemented easily and can increase the preciseness of estimation.