Accelerating urban street canyon wind flow predictions with deep learning method

作     者：Wai-Chi Cheng Tzung-May Fu Wai-Chi Cheng;Tzung-May Fu

作者机构：Shenzhen Key Laboratory of Precision Measurement and Early Warning Technology for Urban Environmental Health RisksSchool of Environmental Science and EngineeringSouthern University of Science and TechnologyShenzhenGuangdong518055China Guangdong Provincial Observation and Research Station for Coastal Atmosphere and Climate of the Greater Bay AreaSchool of Environmental Science and EngineeringSouthern University of Science and TechnologyShenzhenGuangdong518055China National Center for Applied MathematicsShenzhen(NCAMS)ShenzhenGuangdong518055China 

出 版 物：《Building Simulation》 (建筑模拟(英文))

年 卷 期：2025年第18卷第4期

页      面：923-936页

核心收录：

学科分类：08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：supported by the National Natural Science Foundation of China(42375193,42325504) the National Key Research and Development Program of China(2023YFC3706205) the Shenzhen Key Laboratory of Precision Measurement and Early Warning Technology for Urban Environmental Health Risks(ZDSYS20220606100604008) the Shenzhen Science and Technology Program(KQTD20210811090048025,JCYJ20220818100611024) the Guangdong Province Major Talent Program(2019CX01S188) the High-level University Special Fund(G03050K001) 

主　　题：deep learning geometry reading filter large-eddy simulation neural network model urban wind flow prediction urban street canyons 

摘      要：Predicting wind flow statistics in urban areas is important for various environmental and engineering ***,building-resolved computational fluid dynamics(CFD)simulations are the most commonly used and reliable methods to simulate urban wind flows but they are time-consuming which limits their use in real ***,our objective is to develop a surrogate model based on deep learning(DL),which can be used as a faster alternative to CFD methods for urban *** proposed model hypothesis is that the spatial distributions of the time-averaged flow quantities within urban canopies are highly correlated to the local urban *** test this hypothesis,we developed a model to predict the flow in uniform urban street canyons by constructing a geometry reading filter to convert local urban geometry information around the targeted locations into a numerical array as DL model inputs.A standard feedforward DL model is then trained using large-eddy simulation(LES)results to predict the mean wind and turbulence within uniform street *** results show that the model can give fast and accurate predictions compared to LES *** prediction errors are found to range from 5.8%to 36%,and the normalized mean bias magnitudes range from 6.6×10^(−3) to 1.6×10^(−1) for the different flow *** DL model is also found to predict the flow patterns reasonably well,consistent with experimental data similar to the results of coarse-resolution *** model has the potential to be further developed into a robust and practical tool for fast urban flow predictions.