APPLICATION OF NEURAL NETWORKS TO ESTIMATE SOLAR RADIATION I?EATURES FROM AIR TEMPERATURE MEASUREMENTS IN TIZI OUZOU (ALGERIA)

摘要

A model based on neural networks, has been implemented to generate daily sunshine duration data for TiziOuzou (Algeria), knowing air temperature measurements in this location. The neural network chosen in thisstudy, is the hidden layer feedforward network. Its learning rule is based on the backpropagation of the error.Its training basis consists of sunshine duration, minimum and maximum temperature data recorded every dayin Tizi Ouzou from 1990 to 1994. The tests are performed by using daily sunshine duration measurementscollected in this location during 1995. The best estimate of daily sunshine duration is obtained when both themaximum and minimum temperature measurements are used as inputs of the neural network underconsideration.