Research On Haze Objective Forecasts And Correction Technology And Its Application

In recent years, as a main source of urban air pollutant,the haze is continuously striking the heavy industry and developed areas in china, such as Beijing-Tianjin-Hebei region, Yangtze River Delta, and Zhujiang River Delta etc.The frequency of haze occurrence has also increased year by year. The haze is consisted of hundreds of airborne pollutants with complicated composition, some of which can directly penetrate into the human respiration system and cause chronic respiratory diseases. Moreover, the extinction phenomena caused by particulate matters can reduce the air visibility, which will affect traffic safety and result in a great loss. Therefore, research on haze forecast technology has gradually become an important issue for air pollutes control. In this thesis, methods including time series analysis and Karman wave filter are used to improve the modeling of haze forecast.Experimental results validate the efficiency and reliability of the proposed haze forecast technology, which can provide technical insurance and reference for policy decision of government.(1) A correction model based on multivariable stepwise regression algorithm and Karman filter for reliable haze prediction is proposed. The method retrieves the fine grid prediction data from 4 typical monitor stations of ECMWF and converts the grid-based data into station-based data using bilinear interpolation method. The correlations between each prediction factor and prediction variable are established,in order to build the prediction model of visibility. The model is further used for the initial matrix for the Karman filter. Combing with observation data of visibility, the objective forecasts correction model is established. The experimental results show that the accurate rate of the proposed method is greatly improved over the current haze forecast system CUACE used by CMA at present.(2) In order to improve the forecast accuracy and resolve the lagging of prediction and poor prediction precision of time series prediction, a mixed forecast method based on time series analysis and Karman filter is proposed. The stability of time series analysis is first evaluated through graph analysis and eigenvalue analysis(ADF) and the unstable series are converted to stable series via differential algorithm.The model is established based on the stabilized series after transformation, and used as the state function and observation function for the Karman filter. Based on the mixed model of time series analysis and Karman filter,the result of prediction on the aforementioned 4 stations are also values, indicating the effective improvement of prediction precision of the proposed method.