基于深度学习模型的中小河流洪水模拟  

Flood Simulation of Small and Medium-sized Rivers Based on Deep Learning Model

作　　者：张景帅 胡彩虹[2] ZHANG Jing-shuai;HU Cai-hong(College of Hydrology and Water Resources,Hohai University,Nanjing 210098,China;Yellow River Laboratory,Zhengzhou University,Zhengzhou 450001,China)

机构地区：[1]河海大学水文水资源学院,江苏南京210098 [2]郑州大学黄河实验室,河南郑州450001

出　　处：《水电能源科学》2024年第10期11-15,共5页Water Resources and Power

摘　　要：为了研究LSTM模型在我国众多流域中的适用性,在海河、黄河、淮河、长江流域共选取了9个小流域作为研究区域,训练并验证了LSTM在这些流域中的模拟精度。结果表明,随着预见期的增加,LSTM模型模拟效果呈下降趋势,在预见期为1~6h时可以得到良好的模拟结果,在预见期大于6h时,模拟结果较差;随着神经元数量和迭代步数的增加,LSTM模型的模拟效果有所提高,超过一定组合后,模拟效果变化不明显。研究结果可为中小流域的洪水预报提供支持。In order to study the applicability of the LSTM model in many basins,a total of 9 small basins were selected as the research area in the Hai River,Yellow River,Huai River,and Yangtze River basins,and the simulation accuracy of LSTM was trained and verified in these basins.The results indicate that as the forecast periods increase,the simulation performance of the LSTM model shows a decreasing trend.Good simulation results can be obtained when the forecast periods are 1-6 hours,but the simulation results are poor when the forecast periods are greater than 6 hours;As the number of neurons and iteration steps increase,the simulation effect of the LSTM model improves.However,after exceeding a certain combination,the simulation effect does not change obviously.The research results can provide support for flood forecasting in small and medium-sized basins.

关 键 词：深度学习 LSTM 中小河流 洪水模拟 

分 类 号：TV122[水利工程—水文学及水资源]