Research On A Model To Forecast AQI Real-timely

The Air Quality Index(AQI)is an important indicator for measuring the quality of air.How to accurately and quickly forecast AQI in the future has become a hot topic.The emergence of air quality grid real-time monitoring and warning system provides a possibility to forecast AQI in a certain area in real time.However,as the number of samples continues to increase,the calculation of the AQI at next time point is also gradually increased,affecting the real-time performance of the forecast.Although the numerical forecasting model can forecast AQI at multiple time points in the future,due to its low forecasting accuracy,the numerical forecasting model alone cannot meet the public requirements.Aiming at the above problems,the thesis proposes an AQI real-time forecast model.According to different application scenarios,the model is divided into two forecasters in the real-time processing module:single-point AQI real-time forecaster and multi-point AQI real-time forecaster.To solve the problem of poor real-time performance of AQI in forecasting the next time point,a single-point AQI real-time forecaster is established in this thesis based on the calculation characteristics of AQI and the basic idea of KNN algorithm.the meteorological data and air quality data of Beijing during last four years are selected to build a sample set.The value of K is determined by cross-validation method.Then,the distributed extension of KNN algorithm is carried out using Storm distributed flow data processing framework,so that the forecaster can forecast the next time point's AQI in real time.Considering the problem of low accuracy of AQI forecast applied with numerical forecast model,a combined forecast algorithm KNN-KF based on KNN and Kalman filter is proposed in this thesis.Then use the recursive equation of the Kalman filter algorithm to revise correction coefficient in the equation obtained by the initial value of the state vector,and apply meteorological and air quality numerical forecast data and the KNN algorithm to compute the middle forecasting value.Finally construct a KNN-KF combined with algorithm based on KNN and Kalman filtering.Moreover,this thesis also adds wind factors to the system state,forecast factors and correction factors to further reduce the forecast errors.Finally,conduct a distributed extension of the combined KNN-KF algorithm with Storm to establish a high-accuracy multi-point AQI real-time forecaster.In this thesis,the accuracy and performance of single-point AQI real-time forecaster and multi-point AQI real-time forecaster are verified by experiments,and the ideal experimental results are obtained.