主办单位:中国气象局沈阳大气环境研究所
国际刊号:ISSN 1673-503X
国内刊号:CN 21-1531/P

气象与环境学报 ›› 2023, Vol. 39 ›› Issue (4): 31-37.doi: 10.3969/j.issn.1673-503X.2023.04.005

• 论文 • 上一篇    下一篇

基于相似卡尔曼滤波的安徽省WRF模式风速预报订正

吴迪1(),田宏强1,刘辉1,王京景1,左晨亮2,徐晶晶3,*()   

  1. 1. 国家电网安徽省电力有限公司, 安徽合肥 230000
    2. 安徽继远软件有限公司, 安徽合肥 230000
    3. 中国科学院大气物理研究所, 北京 100029
  • 收稿日期:2021-10-29 出版日期:2023-08-28 发布日期:2023-09-23
  • 通讯作者: 徐晶晶 E-mail:wud2734@ah.sgcc.com.cn;xujingjing@mail.iap.ac.cn
  • 作者简介:吴迪, 男, 1972年生, 高级工程师, 主要从事电力气象预警、电网自动电压控制、电网稳定机理等方面的研究, E-mail: wud2734@ah.sgcc.com.cn
  • 基金资助:
    国家电网安徽省电力有限公司科技项目“数值天气预报与电力系统观测源同化技术研究与应用”(B31200200004)

Bias correction of wind speed forecasts for the WRF model in Anhui province based on the analog Kalman filter method

Di WU1(),Hongqiang TIAN1,Hui LIU1,Jingjing WANG1,Chenliang ZUO2,Jingjing XU3,*()   

  1. 1. State Grid Anhui Electric Power Corporation of China (SGCC), Hefei 230000, China
    2. Anhui Jiyuan Software Corporation, Hefei 230000, China
    3. Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China
  • Received:2021-10-29 Online:2023-08-28 Published:2023-09-23
  • Contact: Jingjing XU E-mail:wud2734@ah.sgcc.com.cn;xujingjing@mail.iap.ac.cn

摘要:

应用改进的相似卡尔曼滤波方法对2020年4—12月安徽省19个基本气象站WRF模式预报的10 m风速进行误差订正。结果表明:订正后的风速预报平均偏差从1.35 m·s-1降低至0.08 m·s-1,基本消除了模式的系统误差;均方根误差从1.77 m·s-1减小至0.81 m·s-1。平均风速为3 m·s-1以上的较大风,风速预报的均方根误差从2.01 m·s-1降低至1.19 m·s-1,表明该方法不仅可以有效减小模式的系统误差,还可以大幅减小模式的随机误差。相似卡尔曼滤波可以对无法精确模拟物理过程的数值模式进行误差订正,提高模式在天气系统剧烈变化时的预报准确率,适用于气象要素24~72 h的连续预报。

关键词: 数值预报, 误差订正, 系统误差

Abstract:

The error correction of the 10 m wind speed forecasted by the WRF model at 19 meteorological stations in Anhui province from April to December 2020 was carried out using the improved similar Kalman filter method.The results show that the average deviation of wind speed forecast is reduced from 1.35 m·s-1 to 0.08 m·s-1, which is about an elimination of the systematic error of the model.The root-mean-square error (RMSE) decreases from 1.77 m·s-1 to 0.81 m·s-1.When the average wind speed is more than 3 m·s-1, the root-mean-square error of the wind speed forecast is reduced from 2.01 m·s-1 to 1.19 m·s-1, indicating that this method can not only effectively reduce the systematic error of the model, but also greatly reduce the random error of the model.The similar Kalman filter can correct the error of the physical process model which cannot be simulated accurately, improve the forecast accuracy of the model when the weather system changes dramatically and is suitable for the continuous forecast of meteorological elements for 24~72 hours.

Key words: Numerical weather prediction, Bias correction, Systematic error

中图分类号: