强化学习驱动的城市智能环境管理
《基于强化学习的自适应气候适应策略：纽约曼哈顿海岸洪水风险管理应用》
Kairui Feng, Ning Lin, et al.
Reinforcement learning reduces coastal flood adaptation costs by up to 77%. 强化学习可将沿海洪水适应成本最多降低77%。
RL integrates protection, accommodation, and retreat strategies for higher efficiency. 强化学习整合防护、适应与退避策略，提升经济效益。
RL effectively controls tail risks and adapts to climate uncertainties. 强化学习有效控制尾部风险，适应气候不确定性。
Nature Communications
Tropical cyclone-blackout-heatwave compound hazard resilience in a changing climate
《气候变化下台风-停电-热浪复合灾害韧性研究》
Kairui Feng, Min Ouyang, Ning Lin
Developed a pioneering model to project power outage resilience under evolving climate conditions. 构建了创新模型，预测气候变化背景下电网停电韧性。
Proposed a cost-effective strategy to strengthen power systems against compound climate risks. 提出低成本强化电力系统应对复合气候灾害的策略。
NeurIPS
RainNet: A large-scale imagery dataset and benchmark for spatial precipitation downscaling
《RainNet：空间降尺度降水图像数据集与基准》
Xuanhong Chen, Kairui Feng (co-first), et al.
Compiled a comprehensive real-world precipitation dataset (1979–2018). 收集并整理覆盖1979–2018年的大规模实测降水数据集。
Evaluated multiple downscaling methods to accurately model high-resolution rainfall patterns. 系统评估多种降尺度方法，提升高分辨率降水预测精度。
Provided high-resolution (4km) rainfall dataset and benchmark tools. 公布高分辨率（4km）降水数据及评估基准。
Download 数据集: Google Drive | 代码仓库: Github.