《计算机应用》唯一官方网站 ›› 2026, Vol. 46 ›› Issue (10): 3354-3364.DOI: 10.11772/j.issn.1001-9081.2025091135

• 前沿与综合应用 • 上一篇    

深度学习在气象数据同化中的应用演进与展望

吕海燕(), 周立军, 张杰, 赵媛, 王丽娜   

  1. 海军航空大学 航空基础学院,山东 烟台 264001
  • 收稿日期:2025-09-26 修回日期:2025-11-20 接受日期:2025-11-21 发布日期:2025-12-04 出版日期:2026-10-10
  • 通讯作者: 吕海燕
  • 作者简介:周立军(1982—),男,湖南岳阳人,教授,硕士,CCF会员,主要研究方向:计算智能、航空气象数据同化
    张杰(1984—),男,山东日照人,副教授,硕士,主要研究方向:计算智能
    赵媛(1983—),女,山东临沂人,副教授,硕士,主要研究方向:计算智能
    王丽娜(1983—),女,山东烟台人,副教授,硕士,主要研究方向:计算智能。
  • 基金资助:
    军内科研项目(2025-JHHDDG-F1021)

Evolution and prospect of deep learning in meteorological data assimilation

Haiyan LYU(), Lijun ZHOU, Jie ZHANG, Yuan ZHAO, Lina WANG   

  1. School of Basic Aeronautics,Naval Aviation University,Yantai Shandong 264001,China
  • Received:2025-09-26 Revised:2025-11-20 Accepted:2025-11-21 Online:2025-12-04 Published:2026-10-10
  • Contact: Haiyan LYU
  • About author:ZHOU Lijun, born in 1982, M. S., professor. His research interests include computational intelligence, aviation meteorological data assimilation.
    ZHANG Jie, born in 1984, M. S., associate professor. His research interests include computational intelligence.
    ZHAO Yuan, born in 1983, M. S., associate professor. Her research interests include computational intelligence.
    WANG Lina, born in 1983, M. S., associate professor. Her research interests include computational intelligence.
  • Supported by:
    Military Internal Scientific Research Project of China(2025-JHHDDG-F1021)

摘要:

数据同化(DA)是现代数值天气预报(NWP)系统的核心,但传统方法面临非线性、非高斯假设与计算成本高昂的瓶颈。深度学习(DL)具备强大的非线性拟合与端到端优化能力,为解决上述挑战提供了革命性范式。从计算机科学视角,本文系统梳理DL在气象DA领域的组件替代、范式革新、系统重构三级演进脉络。首先,阐述DL作为代理模型替代观测算子、订正模型误差等传统瓶颈环节的组件替代阶段;其次,剖析以变分自编码器(VAE)和扩散模型为代表的生成式模型,将同化问题重构为概率性生成任务,实现范式革新;进而论述以FengWu、FuXi等人工智能(AI)气象大模型为代表的系统重构阶段,将同化模块深度内嵌,构建同化?预报一体化闭环系统;最后,从物理一致性、数据依赖性、业务部署可行性等方面探讨该领域面临的核心挑战,并展望神经符号融合、开放基准构建等未来研究方向。

关键词: 深度学习, 数据同化, 数值天气预报, 生成式模型, 人工智能气象大模型, 代理模型

Abstract:

Data Assimilation (DA) is the core of modern Numerical Weather Prediction (NWP) systems. However, traditional methods are constrained by bottlenecks such as non-linear and non-Gaussian assumptions as well as high computational cost. Deep Learning (DL), with its powerful non-linear fitting and end-to-end optimization capabilities, offers a revolutionary paradigm to address these challenges. Therefore, from a computer science perspective, this paper reviewed systematically and analyzed a three-stage evolutionary trajectory for the application of DL in meteorological DA: component replacement, paradigm shift, and system reconfiguration. First, it elaborated on the component replacement stage, in which DL serves as surrogate models to replace observation operators and correct model errors, thereby addressing traditional bottlenecks. Second, it provided an in-depth analysis of the paradigm shift stage, where generative models, represented by Variational Auto Encoders (VAEs) and diffusion models, reconstruct the assimilation problem as a probabilistic generation task. Third, it discussed the system reconfiguration stage, in which Artificial Intelligence (AI) meteorological foundation models such as FengWu and FuXi were adopted, and the assimilation module was embedded deeply, so as to construct a closed-loop system of integrated assimilation-forecast. Finally, the paper explored the core challenges in this field, including physical consistency, data dependency, and operational deployment, and prospected future research directions, such as neuro-symbolic fusion and open benchmark construction.

Key words: Deep Learning (DL), Data Assimilation (DA), Numerical Weather Prediction (NWP), generative model, Artificial Intelligence (AI) meteorological foundation model, surrogate model

中图分类号: