大语言模型在民航空管气象中的应用研究

Research on the Application of Large Language Models in Civil Aviation Air Traffic Control Meteorology

摘要: 2025年初开源大语言模型(如DeepSeek)快速发展后，民航新疆空管局依托本地硬件，开展其在空管气象业务中的应用验证，半年内初步构建含气候数据统计、自动化代码生成等功能的业务应用系统。研究完成DeepSeek-7B/14B、阿里通义千问Qwen3-30B等模型的本地化部署，开发气象业务知识智能体、编程助手、无代码气候数据统计工具三类应用。最后从空管业务智能体、多技术协同、数据治理维度提出应用建议，为大语言模型在空管领域落地提供参考。

Abstract: After the rapid development of open-source large language models (such as DeepSeek) in early 2025, the Xinjiang Air Traffic Management Bureau of Civil Aviation, relying on local hardware, carried out application verification of them in air traffic control meteorological business. Within half a year, it initially built a business application system with functions such as climate data statistics and automated code generation. The research has completed the local deployment of models such as DeepSeek-7B/14B and Alibaba Tongyi Qianwen Qwen3-30B, and developed three types of applications: meteorological business knowledge agents, programming assistants, and no-code climate data statistics tools. Finally, application suggestions are put forward from the dimensions of air traffic control business agent, multi-technology collaboration, and data governance, providing references for the implementation of large language models in the field of air traffic control.