Air Quality Warning In Ningbo City Based On Spatio-temporal Data

In recent years,our country perfect management plan continuously on the air pollution.In 2013,state council put forward “article ten of the atmosphere” and rectification in 2014,then listen to the corresponding feedback report and review the modifications in further,implement the laws in 2016.So the environmental air quality forecast is particularly important.On one hand,the theoretical basis can be provided to the national to control air pol ution;on the other hand,it helps people understand the change trend of pollutants and air quality in the future.We can arrange reasonably travel activities in daily life and completes the protective measures.In order to reduce the harm of pollution,this paper would establish the related prediction model and developing the air quality forecast system of Ningbo city based on the Support Vector Machine(SVM).The several aspects in the paper are as fol ows:(1)Actual data of Pollutants preprocessing.according to the relevant provisions of the original sample data to be cleaned.Using partial correlation coefficients to analyze the data between various pollutants and meteorological,then successively ensure the input and output variables of the forecast model.It divided the sample data into the training sample and testing sample using cross validation thought,then trained the training sample to obtain the forecast model and validate the testing sample.This paper divided the testing sample into spring,summer,autumn,winter by analyzing AQI clustering in 2016,then view the forecast effect of the forecast model in different seasons.(2)The establishment of air quality forecast model.It combined the thought of clustering and SVM and combined respectively the SOM neural network and fuzzy C-means clustering algorithm with SVM to establish the forecast model.First,we cluster the training sample,then train each category about SVM model,so as to get a different SVR forecast model,cluster the testing sample.Then,forecast the results in the corresponding respectively on the SVM forecast model.FCM-SVR forecast model precision in the same season is the highest,and the single SVR forecast model precision is the lowest.The summer forecast effect is best and winter forecast effect is worst in the same model.The reason is that the gas is not easy to cycle with the low temperature in winter,weak self-purification ability and the pollutants with,even stop up the certain area and can't scatter for a long time.(3)The correction of the forecast model.This paper separately correct the forecast results considering the influence factors of wind speed and direction.Using the winter model to establish different correction model.According to different direction like east,west,north and south.The results proved that the accuracy of corrected FCM-SVR model is higher than the former.(4)The development of air quality forecast system in Ningbo city.The technologies this system used mainly include front-end,database and server.The main function of Front-end is advance the experience degree of users.The interface design mainly includes Ningbo city air quality forecast results and main pollution and make it as charts to show the change trend of pol utant in nearly one week and twenty-four hours.In addition,we would use map mode to offer the spatial location in eight sites including Ningbo environmentally testing center,Dongqian Lake,etc.The data is stored in the MySQL database.Using ASP.net C# development to carry out the relevant operations for database.The web server is the IIS server.