Research On Urban AQI Forecasting Method Based On Meteorological Factors Analysis

With the development of science and technology,industry and transportation and other industries emit large amounts of pollutants into the atmosphere and leading to air pollution.The Air Quality Index(AQI)is an indicator of air quality.The higher the air quality index is,the worse the air quality will be and the more serious the pollution be caused.When air pollution is serious,it will cause great harm to human health and affect people's lives and travel.At present,the problem of air pollution in China is still grim,especially the winter heating season in the northern part of China.High-precision air quality index prediction is as important as weather forecasting.People can arrange their travel and transportation methods and equipment according to the predicted results,so as to better protect their own health.Considering that AQI has periodic characteristics of that it is high in winter and low in summer,it is considered that there is local similarity between AQI data.Therefore,this paper proposes a K-nearest neighbor method to predict urban AQI and analyze the influence of wind factor on AQI.The effect of the factor is used to corrects K-nearest neighbor's(KNN)prediction results.The experimental results show that the introduction of wind factors has a positive effect on the prediction of AQI by KNN method.However,the influence factors of AQI are complicated,and the KNN method considering the wind factor only is not very accurate for the prediction of the sudden change point.Considering the complex factors of AQI,this paper selects multiple meteorological factors and historical AQIs as the influencing factors to predict AQI,and introduces principal component analysis(PCA)method to reduce the dimension and remove the noise to reduce the computational complexity of the model and improve the calculation rate of the model.The calculation time of the model is reduced,and the meteorological factor data processed by the PCA and the historical AQI are collectively used as the input of the least squares support vector machine(LSSVM),and the AQI of the day is used as the output of the LSSVM to train the LSSVM model.Then using the trained model to predict Single-step and multi-step AQI.The experimental results show that the single-step prediction method based on PCA and LSSVM has better prediction results for AQI than the KNN prediction method considering wind factor,And it is slightlybetter than LSSVM prediction method.And the prediction effect is best when the historical AQI days are 2.The MAPE and RMSE of 1-4 days AQI's prediction results in Nanchang,Nanjing and Hefei cities based on PCA and LSSVM multi-step prediction methods are smaller than the ARIMA method,and the promotion is stronger.