Solar irradiance forecasting model using Artificial Neural Network (ANN)

This thesis presents the Artificial Neural Network (ANN) model for predicting solar irradiance. The inputs of the ANN are the solar irradiance values for previous five one-minute intervals while the output of the ANN is the solar irradiance of the sixthminute interval. The solar irradiance data were obtained from a weather monitoring station at Green Energy Research Centre (GERC) at Universiti Teknologi MARA, Malaysia. During testing, the ANN produced a low mean absolute percentage error (MAPE) of 10.5796% and high coefficient of determination, R2 of 0.8925. The results obtained also show that low MAPE value and high R2 value had shown that the ANN model has a good predictive performance and was useful in predicting solar irradiance.