A hybrid model for online prediction of PM2:5 concentration: A case study

The present study aims to develop a model to predict the daily average concentration of particulate matter with a diameter of less than 2.5 micrometers (PM2:5). This model employed Weather Research and Forecasting (WRF) meteorological model, Monte Carlo simulation, wavelet transform, and Multi-Layer Perceptron (MLP) neural networks. In particular, the MLP and wavelet transformation were combined for prediction purposes. The WRF meteorological model was employed to predict such input parameters of the model as wind speed, wind direction, temperature, rainfall, and temperature inversion. Finally, to achieve a more accurate prediction considering the uncertainty associated with the input data, Monte Carlo simulation was employed. Further, to assess the eectiveness of the proposed model in the real world, it was run in the online mode for 35 days. Numerical results revealed that the combined model yielded acceptable accuracy when using widely used measures. To be specic, according to R measurements, this accuracy was measured as 0.831 over a set of test instances.