Estimation of Hourly Near Infrared Radiation Using Artificial Neural Network and Performance Comparison with the Semi-Empirical Model at Nakhon Pathom Province

Abstract

In this research, methods for estimating near infrared radiation (NIR: 0.695–2.8 micron) at Silpakorn University, Nakhon Pathom province (13.82◦N, 100.04◦E) have been developed using an artificial neural network (ANN) and a semi-empirical model. The input data of these models consist of aerosol optical depth (AOD) and precipitable water (W) measured by a Sunphotometer, and clearness index (kt) obtained from ratio of measured incident solar radiation to calculated extraterrestrial solar radiation. The ANN and semi-empirical models were formulated using the collected data at Nakhon Pathom station for the period of 2009-2015. Then, the results obtained from these models were tested and validated against the measured data at the station during a two-year-period (2016-2017). The comparison results show that the near infrared radiation obtained from the ANN and semi-empirical models are in reasonable agreement with the measurement. The root mean square difference (RMSD) are 6.08% and 4.47%, and the mean bias difference (MBD) are 4.91% and 3.02% for the ANN and semi-empirical models, respectively. Keywords: near infrared radiation, artificial neural network, semi-empirical model, Nakhon Pathom