Air temperature forecasting using artificial neural networks with delayed exogenous input

Solar systems are influenced by their functioning temperature, which is mainly a result of the ambient air temperature as well as the level of sunlight. For this reason, we have proposed an intelligent autoregressive model based on the multilayer neural networks with delayed exogenous input sequence representing the Global Solar Radiation (GSR) data to forecast the Air Temperature (AT) values on a half hour scale in Marrakesh, Morocco, which is recorded during 2014. The developed model will helps us to monitor and to overcome such a problem related to the intermittence or the absence of AT measurements in this region. For building this model, the database is divided into two subsets containing the same size of data. The first one used to train the model using Levenberg-Marquardt algorithm, and the second is used to test the performance and the robustness of the proposed model. The obtained results shown that the developed model has two hidden layers with a few hidden neurons number can predict accurately the target that is demonstrated by the high value of the correlation coefficient, which reaches 99.80.