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

Abstract

In this paper, we aim at developing a model to predict the daily average concentration of particulate matters with a diameter of less than 2.5 micrometers (PM 2.5 ). In the introduced model, we incorporate Weather Research and Forecasting (WRF) meteorological model, Monte Carlo simulation, wavelet transform, and multilayer perceptron (MLP) neural networks. In particular, the MLP and wavelet transformation are combined for prediction. In order to predict the model's input parameters, including wind speed, wind direction, temperature, rainfall, and temperature inversion, the WRF meteorological model is used. Finally, according to the available uncertainty in the input data and in order to achieve a more accurate prediction, the Monte Carlo simulation is utilized. In order to assess the effectiveness of the model in the real world, it has been conducted in an online mode for 35 days. Numerical results give an acceptable accuracy in terms of some widely used measures. In particular, taking into account the R measurements, it is equal to 0.831 over the set of test instances.