Research And Application Of Fine-Grained Air Quality Prediction And Inference Model Based On Urban Big Data

With the development of urbanization and the expansion of city scale,the car usage has also risen sharply,and air pollutant emissions from factories,which are reasons for deterioration of the urban environment.Thus,the air quality of cities has gradually become a hot topic for urban residents.Now it’s an urgent need to use historical urban air quality data,meteorological data,POI data,weather forecast data and other urban big data to design an air quality prediction and inference models,which can help residents plan outdoor arrangements and activity.In the previous researches,the most of works were usually solved from a single time dimension.About this issue,our work starts from the time and space dimensions,considering the continuity of air pollutants in time and the diffusivity in space.In order to achieve better prediction,we combine the prediction results from two dimensions with multiple factors.Due to few air quality monitoring stations in the city and no monitoring data in a large number of areas,this thesis also proposes a model for inferring air quality.First of all,this thesis builds the Air QP-DNN model for the air quality stations distributed in the city,combined with AQI historical data,meteorological data,and weather forecast data.The Air QPDNN model predicts the AQI value from the time dimension and the space dimension,and obtains the corresponding predicted value.Then,according to the current weather condition,combining the prediction results of the two dimensions to further improve the accuracy of the predicted value.Secondly,under the condition with the limited AQI data of monitoring stations in the city,the air quality of areas without air monitoring station in the city is inferred by combining with the POI data,road network data,and traffic flow data.Different from the previous spatial interpolation algorithm,we apply the similarity theory of the Third Law of Geography to the spatial distribution of air quality.the air quality inference model based on the Third Law of Geography was built by selecting data from multiple fields and calculating similarity degree,and the accuracy of the inference has been improved effectively.Finally,this thesis uses historical air quality data,meteorological data,weather forecast data,POI data,road network data,and traffic flow data to verify the accuracy of the models.With the Air QPDNN and the air quality inference model based on the Third Law of Geography,multiple routes are planned by the user’s origin-destination,which can select best option with a shorter length route with better air quality.