developing a model for ozon concentration prediction using artificial neural network

ozone concentration in metropolitan areas frequently exceed regulatory standards now. since it is harmful on human health and environment, its modeling and visualization are of vital important. modeling and prediction of ozone is required by urban managers to control and prevent its effects. in this paper, important parameters influencing the hourly ozone concentration are estimated by using data collected in azadi and imam khomeini air quality stations in 2009-2010. in this context, the correlations between ozone concentration and meteorological parameters such as relative humidity, temperature, pressure, wind speed, and its directions are determined by linear regression and principle component analysis. results showed that ozone concentration is mostly affected by relative humidity, temperature and wind speed. the effects of relative humidity and temperature on ozone concentration are attributed to photochemical processes. while the correlation between the ozone and wind speed is due to the ozone transfer from nearby regions. after determining the important parameters, a neural network was used for forecasting of ozone concentration of the next 24 hours for a week in four different seasons. the input to the neural network model was relative humidity, temperature and wind speed. results of the implementation showed that the model can predict the ozone concentration in azadi and imam khomeini air quality stations by an accuracy of 67 to 97 percent for the next 24 hours. urban managers for more efficient management and control of the ozone concentration can use results of this research work.