Multiscale Recurrence Plot Analysis And Prediction Research On Air Quality Data

Currently,air pollution has become one of the environmental and social issues of growing public concern in China.Air pollution not only brings a serious burden to the ecological environment but also has a great negative impact on the physical and mental health of human beings.Accurate analysis and prediction of air quality have become an important measure to control air pollution and improve air quality.Taking Beijing multi-site air quality data as the research object,aiming at the characteristics of air quality data,the dynamic correlation of air quality influencing factors is explored by recurrence plot,and the air quality time series are modeled by the deep neural network.Specifically,the following three aspects are included.First,to fully explore the potential high-dimensional correlation information among air quality factors and assess the influence degree of factors on air quality,a multi-scale recurrence analysis model for air quality based on cross recurrence plot is proposed.Structurally,scaling pre-processing is used to divide air quality data,and cross recurrence plot and recurrence quantitative analysis are used to provide qualitative and quantitative explanations for the dynamic correlations among air quality factors at multiple scales.The simulation results show that this method can visualize and quantify the dependence among air quality factors from multiple scales.Second,considering that air quality is influenced by multiple factors,a multi-time scale prediction method for air quality data is proposed.The method is oriented to the single-site pattern and performs multi-timescale dynamic recurrence analysis and prediction for air quality data.In structure,multi-time scale recurrence dynamic analysis is applied to mine the main factors affecting air quality and provide high-quality input conditions for prediction.Further,a deep learning model is used to model the air quality data.Simulation results show that the method has good nonlinear fitting performance compared with models such as SVM,DBN,and LSTM.Finally,for the problem that air quality factors are influenced by spatial and temporal scales,a air quality spatial-temporal scale prediction method is proposed.The method adopts a multi-site collaborative pattern to analyze and model the air quality data with spatial-temporal correlation.Architecturally,the spatial-temporal scale dynamic analysis is used to explore the mechanism of factors in the multi-site on air quality.Further,a deep learning model is used as the basis,and a teaching-learning-based optimization algorithm is introduced for parameter configuration.The effectiveness of the method is verified through experimental simulations,using multiple metrics evaluation and statistical analysis.Figure 20;Table 14;Reference 61...