系统仿真学报 ›› 2018, Vol. 30 ›› Issue (4): 1482-1489.doi: 10.16182/j.issn1004731x.joss.201804033

• 仿真应用工程 • 上一篇    下一篇

多元逐步回归与卡尔曼滤波法在霾预报中应用

咸云浩1, 张恒德2, 谢永华1,3, 杨乐1   

  1. 1.南京信息工程大学 计算机与软件学院,南京 210044;
    2.中国气象局 国家气象中心,北京 100081;
    3.南京信息工程大学 江苏省网络监控中心,南京 210044
  • 收稿日期:2016-05-12 修回日期:2016-07-11 出版日期:2018-04-08 发布日期:2019-01-04
  • 作者简介:咸云浩(1991-),男,江苏淮安,硕士,研究方向为人工智能、机器学习、污染物预报。
  • 基金资助:
    国家自然科学基金(61375030),科技部大气污染专项(JFY2016ZY01002213)

Application of Multiple-Stepwise and Kalman Filtering in Haze Forecast

Xian Yunhao1, Zhang Hengde2, Xie Yonghua1,3, Yang Le1   

  1. 1.School of Computer and Software, Nanjing University of Information science and Technology, Nanjing 210044, China;
    2.Nation Meteorological Center of CMA, Beijing 100081, China;
    3.Jiangsu Engineering Center of Network Monitoring, Nanjing University of Information and Technology, Nanjing 210044, China
  • Received:2016-05-12 Revised:2016-07-11 Online:2018-04-08 Published:2019-01-04

摘要: 针对目前霾预报的重要性和霾客观预报准确率较低,提出基于多元逐步回归算法和卡尔曼滤波算法的霾客观预报订正技术。利用多元逐步回归法控制因变量的物理因子,建立能见度预报方程,利用卡尔曼滤波法根据实况资料对多元逐步回归算法中回归系数进行订正,建立霾客观预报订正模型。以北京站、广州站、南京站、杭州站四个站为例,对站点进行预报实验和检验。实验结果表明,与业务上运行的雾-霾数值预报系统(CUACE)进行对比,提出的多元逐步回归与卡尔曼滤波法的预报准确率有所提高。

关键词: 多元逐步回归, 卡尔曼滤波, 预报模型, 能见度, 霾预报

Abstract: Considering the importance of objective haze forecast and the low accuracy rate of haze forecast, a new haze objective forecast correction method based on the multiple stepwise regression algorithms and the Kalman filtering algorithm is proposed. The multiple stepwise regression method is used to control the physical factor of the dependent variable, and the visibility forecast equation is established. The Kalman filtering method is adopted to correct the regression coefficient in multivariate stepwise regression algorithm according to the actual data, and the haze objective forecast correction model is established. The experiments are carried out in Beijing, Guangzhou, Nanjing and Hangzhou. The experimental results show that comparing with the operational running fog - haze numerical prediction system (CUACE), the prediction accuracy of the multiple stepwise regression and Kalman filtering method is improved.

Key words: multiple-stepwise regression, Kalman filtering, forecasting model, visibility, haze forecast

中图分类号: