基于3D卷积神经网络的热带气旋强度估测  

Estimation of Tropical Cyclone Intensity Based on 3D Convolutional Neural Networks

作　　者：王瑜 孙凤远 WANG Yu;SUN Fengyuan(Hunan Modern Logistics College,Changsha 410131,China;Unit 75841 of PLA,Changsha 410000,China)

机构地区：[1]湖南现代物流职业技术学院,湖南长沙410131 [2]中国人民解放军75841部队,湖南长沙410000

出　　处：《现代信息科技》2025年第5期51-55,61,共6页Modern Information Technology

摘　　要：在气象学和灾害管理领域,热带气旋(Tropical Cyclone,TC)的强度估测具有至关重要的意义。随着科技的进步,基于深度学习的方法在热带气旋强度估测上展现出了卓越的性能,为相关领域的研究和实践提供了有力支持。文章聚焦于热带气旋的时空特征,并结合深度学习技术提出了一种创新的TC强度估测方法—Time-space 3D Network(T3D-Net)模型。该模型在TCIR数据集上的MAE为6.92 kt,RMSE为9.14 kt,与现有的多个热带气旋强度估测方法相比,该方法展现出了一定的竞争性和优越性。In the fields of meteorology and disaster management,the estimation of the TC intensity is of vital significance.With the advancement of technology,methods based on Deep Learning have demonstrated excellent performance in the estimation of the TC intensity,providing strong support for research and practice in related fields.Focusing on the spatiotemporal characteristics of TC,combined with Deep Learning technology,this paper proposes an innovative TC intensity estimation method namely the T3D-Net model.The MAE of this model on the TCIR dataset is 6.92 kt,and the RMSE is 9.14 kt.Compared with multiple existing methods for estimating the TC intensity,this method exhibits a certain competitiveness and superiority.

关 键 词：热带气旋强度估测 3D卷积神经网络 TCIR 时空特征 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP391[自动化与计算机技术—控制科学与工程]