WRF Model Based Quantized Kernel Adaptive Filter for PM2.5 Concentration Online Prediction

With increasing attention have been paid to the prevention and treatment of air pollution in China, an accurate model to predict the pollutants concentration is urgently needed. At present, most of the air pollutants prediction methods cannot forecast the PM_2.5 concentration efficiently in real time when facing large-scale data, and few models can simultaneously consider the influences of high and surface meteorological factors on PM_2.5 concentration. In the view of the above two problems, an improved online prediction model for PM_2.5 concentration called Weather Research and Forecasting model based quantized kernel adaptive filter (WRF-QKAF) is proposed in this paper. The WRF-QKAF model can accurately predict PM_2.5 concentration by combining WRF model and QKAF algorithm. In addition to the air pollutant factors and meteorological factors near ground, the stereoscopic meteorological factors generated by the WRF model are also utilized in the WRF-QKAF model. By this way, the WRF-QKAF model can fully consider the different meteorological conditions and atmospheric factors, and can simulate the physical and chemical transmission process of PM_2.5 better. Furthermore, the new model can reduce the complexity of modeling and track the time-varying characteristics of meteorological data, which enables high accuracy online prediction of PM_2.5 concentration. Guangzhou, which located in the Pearl River Delta and has serious air pollution problem, is selected as the research area, and the online prediction and analysis of the PM_2.5 concentration in 2016 is presented in this paper. Experiments show that the proposed model can characterize the change of PM_2.5 concentration well, thus increasing the practicability and real-time performance of the model, and achieving better prediction results than the comparison algorithms.