English

运筹学学报

• 运筹学 • 上一篇    下一篇

基于时间序列分析的北京地区PM2.5浓度研究

李为东1 李莉1 徐岩1,*   

  1. 1. 北京科技大学信息与计算科学系, 北京 100083
  • 收稿日期:2017-12-08 出版日期:2018-06-15 发布日期:2018-06-15
  • 通讯作者: 徐岩 E-mail: xuyan@ustb.edu.cn
  • 基金资助:

    国家自然科学基金(Nos. 11671032, 11321024), 中央高校基础科研业务费(No. FRF-TP-17-024A2)

The concentration research of PM2.5 in Beijing with time series analysis

LI Weidong1  LI Li1  XU Yan1,*   

  1. 1. Department of Information and Computer Science, University of Science and Technology Beijing, Beijing 100083, China
  • Received:2017-12-08 Online:2018-06-15 Published:2018-06-15

PDF

302

摘要：

基于中国环境监测总站公布的实时空气质量监测数据, 利用时间序列模型对PM2.5指标的数据进行了平稳性、纯随机性检验, 同时进行了模型阶数、未知参数估计以及模型显著性检验与优化. 最终在此基础上建立了指标预测的数学模型, 并对未来三天的PM2.5浓度值进行预测. 进一步地, 基于向量自回归(VAR)模型, 对北京市万寿西宫站PM2.5数据进行相关性分析, 研究空气中污染物O_{2}、NO_{2}、CO、O_{3}、PM10与PM2.5的动态影响关系. 研究发现当天的PM2.5浓度会受到前几天PM2.5、PM10、O_{3}、SO_{2}等污染物浓度的影响, 其中PM10对PM2.5的影响最为明显且持续时间最长, O_{3}、SO_{2}对PM2.5浓度的影响在二、三期最为明显.

关键词: PM2.5, 时间序列, VAR模型

Abstract:

We have analyzed the main components of PM2.5 with time series analysis based on the data which were collected in Beijing city from the http://www.cnemc.cn/. We have processed the stationarity, pure randomness of the data and the orders, parameters and significance testing of the model. Meanwhile we also predicted the PM2.5 concentration based on the rules which were got from the data. The results of prediction have showed that the error was in the reasonable range. We also analyzed the PM2.5 relative analysis and explored the dynamic relation of PM2.5 with other pollutants such as SO_{2}、NO_{2}、CO、O_{3}、PM10 on the Wanshougong position through vector auto-regressive model (VAR). It has been illustrated that the concentration of PM2.5 could be influenced by SO_{2}、NO_{2}、CO、O_{3} and PM10 with the other days. We found that PM2.5 was influenced mostly by PM10 with a long time. The influence of O_{3} and SO_{2} to PM2.5 concentration was remarkable at the second and third periods.

Key words: PM2.5, time series analysis, vector auto-regressive model