基于机器学习模型的平原水网区H-ADCP流量测算精度比较研究  

Accuracy comparative study on H-ADCP flow measurement based on machine learning models in plain water network area

作　　者：潘道宏[1] 辛朋磊[2] 夏飞[1] 王琪[1] 李昂 查红[1] PAN Daohong;XIN Penglei;XIA Fei;WANG Qi;LI Ang;ZHA Hong(Yancheng Substation,Bureau of Hydrology and Water Resources Survey of Jiangsu Province,Yancheng 224000,China;Nantong Substation,Bureau of Hydrology and Water Resources Survey of Jiangsu Province,Yancheng 226001,China)

机构地区：[1]江苏省水文水资源勘测局盐城分局,江苏盐城224000 [2]江苏省水文水资源勘测局南通分局,江苏南通226001

出　　处：《水利水电快报》2025年第2期22-25,共4页Express Water Resources & Hydropower Information

摘　　要：为提升H-ADCP在平原水网区的测流精度,以淮河流域平原水网区东台(泰)水文站为例,选取缆道和固定式H-ADCP 2022~2023年河道全断面点流速实测数据,基于Python软件,分别运用6种机器学习模型:多层感知机模型、支持向量机回归模型、最小二乘线性回归模型、岭回归模型、袋装算法和随机森林算法拟合河道断面流量,并对6种模型的断面流量计算精度进行了比较分析。结果表明:随机森林算法的流量计算精度高于其他模型,系统误差、随机不确定度、符号检验、适线检验、数值检验均能达到水文资料整编规范三类精度要求。研究成果对H-ADCP流量在线监测的应用推广有借鉴意义。In order to improve the accuracy of H-ADCP flow measurement in plain water network,we took the Dongtai hydrographic station of Huaihe River Basin plain water network area as an example,and selected full section flow velocity measurement data of the cable type and fixed H-ADCP from 2022 to 2023.Then we used six machine learning models,namely multi-layer perception,support vector regression machine,least squares linear regression,ridge regression model,bagging method,and random forest algorithm to fit the river section flow based on Python software.And an comparative study on accuracy of the flow by the 6 models was conducted.The results showed that the random forest algorithm had a higher accuracy than other models,and the system error,random uncertainty,symbol test,fitting test,and numerical test could meet the third class accuracy requirements of the hydrological data compilation standards.The research results can provide a reference for the application and promotion of H-ADCP online traffic monitoring.

关 键 词：流量测算 H-ADCP 机器学习模型 平原水网区 淮河流域 

分 类 号：P332.4[天文地球—水文科学]