ARTIFICIAL NEURAL NETWORKS FOR PREDICTING GLOBAL SOLAR RADIATION: CASE STUDY AL-AIN CITY (UAE)

摘要

Measured air temperature, relative humidity, wind and sunshine duration measurements between 1995 and 2007 for Al Ain city in United Arab Emirates (UAE) were used for the estimation of monthly average daily global radiation on horizontal using Artificial Neural Network technique. Weather data between 1995 and 2006 were used for training the neural network, while the data of year 2007 was used for validation. The predications of Global Solar Radiation (GSR) were made using four combinations of data sets namely: 1) Sunshine, Temperature, Humidity and wind 2) Sunshine, Temperature and Humidity 3) Sunshine, Temperature and wind 4) Sunshine, wind and Humidity 5) Temperature, Wind and Humidity 6) Sunshine and Temperature 7) Sunshine and Wind 8) Sunshine and Humidity 9) Temperature and Wind 10) Temperature and Humidity. The ANN models with different input parameters have R~2= 0.7928 or higher, RMSE values vary between 0.276 to 0.79963 and small MBE ranging from - 0.0018895 to 0.0000882.