基于BP神经网络和三次样条插值法的感潮河段水位预报  

Water level forecast of tidal river sections based on the BP neural network and cubic spline interpolation method

作　　者：季俊杰[1] 徐瑶瑶 闻昕[3] 纪凯文 马晶洁 JI Junjie;XU Yaoyao;WEN Xin;JI Kaiwen;MA Jingjie(Jiangsu River Course Administration Bureau,Nanjing 210029,China;Water Conservancy Engineering Management Office of Taihu Region of Jiangsu Province,Suzhou 215100,China;College of water Conservancy and Hydropower Engineering,Hohai University,Nanjing 210098,China)

机构地区：[1]江苏省河道管理局,江苏南京210029 [2]江苏省太湖地区水利工程管理处,江苏苏州215100 [3]河海大学水利水电学院,江苏南京210098

出　　处：《江苏水利》2024年第7期33-37,46,共6页Jiangsu Water Resources

基　　金：江苏省水利科技项目(2020065)。

摘　　要：针对常熟水利枢纽外江侧为感潮河段,提出一种基于BP神经网络和三次样条插值结合的感潮河段水位预报方法。结果表明,潮位预报预见期为1 d与2 d时,模型绝对误差分别为0.06 m与0.18 m,合格率分别为87.5%与70.9%,满足《水文情报预报规范》所规定的发布正式预报要求;预见期为3 d时,模型绝对误差与合格率分别为0.28 m与61.4%,满足参考性预报要求。水位预报预见期为1 d时,模型绝对误差为0.07 m,适用于枢纽的精细化调度;预见期为2 d与3 d时,模型绝对误差分别为0.13 m和0.18 m,可为枢纽运行提供精准的外江侧水位预报。A water level prediction method for tidal river sections on the outer river side of Changshu Water Conservancy Hub is proposed based on a combination of BP neural network and cubic spline interpolation.The results show that when the forecast period for tidal level forecast is 1 day and 2 days,the absolute errors of the model are 0.06 m and 0.18 m,respectively,and the qualification rates are 87.5%and 70.9%,meeting the requirements for issuing formal forecasts as stipulated in the“Hydrological Information Forecasting Specification”;When the forecast period is 3 days,the absolute error and qualification rate of the model are 0.28 m and 61.4%,respectively,meeting the reference forecast requirements.When the forecast period for water level is 1 day,the absolute error of the model is 0.07 m,which is suitable for the refined scheduling of the hub;When the forecast period is 2 days and 3 days,the absolute errors of the model are 0.13 m and 0.18 m,respectively,which can provide water level information reference for the operation of the hub.

关 键 词：感潮河段 水位预报 水利枢纽 

分 类 号：TV123[水利工程—水文学及水资源]