Investigation of artificial intelligence approaches (ANN-MLP, CAFIS) for the daily prediction of winter air pollutants (CO₂, SO₂, NO_x, O₃) in Hamedan city using meteorological data

Recently, several factors such as the physical growth of cities and the increased number of industries and cars, Hamedan city in Iran has faced the issue of air pollution. Due to the increased fuel consumption for heating purposes in the cold winters of this city, the pollution rate is higher in this season. Hamedan is surrounded by Alvand Mountains, which makes the air pollution control policies and air pollution management more important in this city. In the present study, the new methods of artificial neural network and meteorological data were used and compared as a tool for the prediction and warning of air pollution in Hamedan city. Highly accurate methods are available for the prediction of meteorological variables, which provide reliable data for the prediction of air pollution. In order to avoid over-training and assess the network compatibility with the lack of data, the minimum number of the data input data was used in this study. According to the results, the combined approaches of the artificial neural network were applicable in this regard, while ANN-MLP with the momentum learning rule and the TanhAxon transfer function yielded more accurate results compared to CAFIS.

Keywords:

Artificial Neural Network , Early Warning , Air Pollutants