Estimation of Daily Global Solar Irradiation in Indonesia with Artificial Neural Network (ANN) Method

Abstract

This paper demonstrates the use of neural network method for estimating daily global solar irradiation in the horizontal surface by meteorological data in Indonesia. The database consists of 1826 daily measured data, in term of sunshine duration, average temperature, average relative humidity and global solar irradiation. The data has been collected in Jakarta (altitude 6°15’S, longitude 106°45’E), a city as capital of Indonesia. The 1461 daily measured data between 2005 and 2008 are used to train the neural networks while the data for 365 days in 2009 are used as testing data. The estimation of global solar irradiation were made using four combinations of proposed model of data sets namely: (i) day of the year, daily average relative humidity and daily average temperature as inputs, (ii) day of the year, daily average relative humidity and sunshine duration as inputs, (iii) day of the year, daily average temperature and sunshine duration as inputs, (iv) day of the year, daily average relative humidity, daily average temperature and daily sunshine duration as inputs. The output for the four combinations proposed model is daily average global solar irradiation. The comparison result show that the best model to estimate global solar irradiation is performed by (iii) and (iv) that obtain the Mean Average Percentage Error (MAPE) at below 10%.