用AI把天气预报精准到每根输电塔，提前预警台风破坏
From AI Weather Prediction to Infrastructure Resilience: A Real-Time Correction-Downscaling Framework for Tropical Cyclone Impact Forecasting
- 分步处理偏差校正与地形细化，避免误差累积
- 500米分辨率风场，站点风速误差降低38.8%
- 适合电力系统、防灾部门做实时风险预警
本文解决基础设施韧性中的关键短板：将快速、全局的AI天气预报转化为资产尺度、可操作的风险情报。提出基于AI的修正-降尺度框架（ACDF），结合实时偏差校正、地形感知降尺度及基于易损性的输电系统风险评估，用于热带气旋影响预测。ACDF分离风暴尺度偏差校正与地形敏感性优化，缓解误差传播，恢复亚千米级风速变化，决定结构荷载。在11个影响浙江的台风上进行留一风暴评估，生成全省尺度500米分辨率风场，相比Pangu-Weather在站点尺度风速均方误差降低38.8%，单次12小时循环仅需约25秒（单块GPU）。以台风黑格比为例，成功再现高风速尾部特征，识别出沿海高风险走廊，并提前标记了后续发生故障的输电线路，实现塔与线路级别的可行动指导。ACDF为从全球AI天气预报到关键基础设施影响型预警提供了端到端路径。
原文摘要 · Abstract (English)
This paper addresses a missing capability in infrastructure resilience: turning fast, global AI weather forecasts into asset-scale, actionable risk intelligence. We introduce the AI-based Correction-Downscaling Framework (ACDF), which combines real-time bias correction, terrain-informed downscaling, and fragility-based power transmission system risk assessment for tropical cyclone impacts. ACDF separates storm-scale bias correction from terrain-aware refinement, mitigating error propagation while restoring the sub-kilometer wind variability that governs structural loading. Tested on 11 typhoons affecting Zhejiang, China under leave-one-storm-out evaluation, ACDF produces 500 m wind fields over a province-scale domain, reduces station-scale wind-speed MAE by 38.8% relative to Pangu-Weather, and runs in approximately 25 s per 12-h cycle on a single GPU. In the Typhoon Hagupit case, ACDF reproduced observed high-wind tails, identified a coastal high-risk corridor, and flagged the transmission line that subsequently failed, demonstrating actionable guidance at tower and line scales. ACDF provides an end-to-end pathway from global AI weather forecasts to operational, impact-based early warning for critical infrastructure.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。