Forecasting PM10 concentrations using neural networks and system for improving air quality

In this paper using Artificial Neural Network (ANN) are presented forecasting results of PM10 concentrations for the city of Sarajevo. Input data of the proposed model are meteorological variables (wind speed, humidity, temperature and pressure) and pollution variable (PM10 concentration) recorded in the Federal Institute for Hydrometeorology from 2010 to 2013. The proposed model is tested on the several cases and the results are satisfactory. The results of the forecasting show the different effects that certain meteorological parameters have on the temporal prediction of concentrations of PM10. It can also be concluded that ANN approach is very useful in terms of the time series forecast the concentrations of PM10 particles with good forecasting results. Also, it is presented the idea of a unified system for air quality improvement, which involves a variety of systemic measures in the areas affected by an increase of PM10 concentrations.