An early warning system to forecast air pollutant concentration

Abstract

The goal of this research is to identify the seasonal fluctuation in concentrations of important air pollutants such as SO2, NO2, NO, CO, O3, PM2.5, PM10. This study conducted a detailed investigation of air pollutants and their interaction with meteorological factors in Delhi, India, from 2019 to 2021. To better understand the relationship between air contaminants and meteorological characteristics, single and multiple parameters are controlled. The study discovered that changing the variable causes a significant shift in the relationship between seasonal and regional attributes. The correlation between the AQI (Air Quality Index) and all of the air contaminants and climatic variables available for the study are explored. An early warning system is proposed, having three modules: Pre-processing, Forecasting and Evaluation module. In Pre-processing to make the time series input data smooth for prediction, Variational Mode Decomposition technique is utilized, followed by prediction of air pollutants. The proposed model used BiLSTM and GRU model for prediction. The results are validated using standard validation parameters RMSE, MAPE, MDA, MAE and MDAPE. The AQI level of last 30 days is also calculated using the standard formula. The predicted AQI level and actual AQI level results have shown a remarkable performance of the proposed model. The proposed model is also validated with the existing model and the lesser error value approves the good performance of the proposed hybrid model over existing one. newline