Research On Environmental Pollution Prediction Based On The Analysis Of Spatial-Temporal Characteristics

With the development of the times and the improvement of environmental awareness,people pay more and more attention to air pollution.A large number of environmental monitoring points have been set up by relevant agencies to effectively measure the concentration of pollutants in the air.These monitoring points provide us with a large number of data support.The analysis and prediction of the measured air pollution concentration data can effectively guide the formulation of air pollution control measures.In order to further improve the accuracy of air pollution concentration prediction,this paper comprehensively analyzes the spatial-temporal characteristics of air pollution data and the correlation between various pollutants,and proposes a new pollution concentration prediction model.The main contents of this paper are as follows:(1)Data preprocessing and correlation analysis.Firstly,the air pollution data and meteorological data are processed with missing value and non-numeric data are quantified.In the preprocessed air pollution data set,the autocorrelation is used to measure the temporal correlation of pollution data,and the Pearson correlation coefficient is used to calculate the spatial correlation between monitoring points and the factor correlation between various pollutants.Correlation analysis provides the data analysis basis for the prediction model.(2)Based on the wind direction cosine adaptive nearest neighbor,the calculation method of spatial correlation degree of environmental monitoring points is proposed.Considering that a variety of pollutants do not exist independently in space,multiple monitoring points will spread and influence each other,and wind direction factors will also affect the diffusion of pollution.The spatial correlation is measured by calculating the similarity of wind direction cosine,and the experimental results show that the prediction accuracy of the air pollution concentration prediction model can be effectively improved by constructing the space-time matrix of wind direction cosine adaptive nearest neighbor,compared with the European space-time matrix of near neighbor and the space-time matrix of all monitoring points.(3)Encoder-Decoder model based on multi-factor attention mechanism.Because the time dimension of air pollution concentration data is relatively sparse,the data show strong volatility,so the prediction effect is not good when only considering the time dependence of pollution data.It is necessary to consider the mutual transformation and cancellation among various pollutants,that is,the correlation between various pollutants and the interaction between multiple monitoring points in space.The multi-factor attention mechanism was introduced and the encoder-decoder model was used to predict the air pollution concentration.The experimental results show that the air pollution concentration prediction model is better.(4)Design and implementation of air pollution concentration analysis system.In this system,Vue and node.js are used to separate the front and back ends,and Baidu map API is used to realize map visualization and Echarts to realize line chart and histogram visualization of pollution data and prediction accuracy.These make the system interface clear and diverse.