多源数据嵌套式同化在水量监测精度提升中的应用  

Application of nested multi-source data assimilation methodology in water quantity monitoring accuracy improvement

作　　者：蒋飞卿 牛智星[1,2,3] 朱易青 王蓓 周安辉 嵇海祥[1,2,3] JIANG Feiqing;NIU Zhixing;ZHU Yiqing;WANG Bei;ZHOU Anhui;JI Haixiang(Nanjing Research Institute of Hydrology and Water Conservation Automation of Ministry of Water Resources,Nanjing 210012,China;Jiangsu NIHWA Technology Co.,Ltd.,Nanjing 210012,China;Research Center on Hydrology&Water Resources Monitoring of MWR,Nanjing 210012,China;Liyang Jiangnan Engineering Testing Co.,Ltd.,Liyang 213300,China;Xishuangbanna Branch Office,Yunnan Hydrological and Water Resources Bureau,Xishuangbanna 666100,China)

机构地区：[1]水利部南京水利水文自动化研究所,江苏南京210012 [2]江苏南水科技有限公司,江苏南京210012 [3]水利部水文水资源监控工程技术研究中心,江苏南京210012 [4]江苏省溧阳市江南工程检测有限公司,江苏溧阳213300 [5]云南省水文水资源局西双版纳分局,云南西双版纳666100

出　　处：《人民长江》2025年第1期81-87,124,共8页Yangtze River

基　　金：国家重点研发计划项目“山洪灾害风险防控区划与全过程监测防范关键技术”(2023YFC3006700);水利部水利科技项目“基于数据同化方法提高澜沧江允景洪站水量监测精度的技术研究”(2023-117);国家自然科学基金重大研究计划项目“西南河流源区水文-环境多过程的立体感知及演变研究”(92047203);水利部南京水利水文自动化研究所科研项目“雷达波测流数据降噪及流量计算优化技术研究”(YJZS0624003)。

摘　　要：数据融合同化可以实现多源观测数据和模型模拟的优势互补,提升监测精度与可靠性。澜沧江下游的水量变化对沿岸国家影响重大,但其水量监测面临着流域复杂、水利工程影响等诸多挑战。为提高水量监测的精度与效率,提出一种耦合水动力模拟的多源数据嵌套式融合同化方法。首先利用人工实测数据构建基于机器学习LASSO模型的侧扫雷达精度提升方案,在此基础上构建河道水动力数值模拟模型,并利用提升后的侧扫雷达监测流速优化水动力模型参数,形成多层级多源数据的嵌套式融合同化,提高水量模拟精度的同时,将点观测数据扩展到全河道,扩展水量要素的获取范围,最后在澜沧江允景洪站进行应用验证。结果表明:基于机器学习LASSO模型的精度提升方案,使侧扫雷达在线监测系统的精度较常规方法提升22.93%;多层级多源数据的嵌套式融合同化模式有效提升了断面流量的模拟精度,验证期相关系数为0.935,并获取了建模河道内任意点的水位、流量、流速等水文要素数据。研究成果可为澜沧江水量监测提供技术支撑。Data fusion and assimilation can complement the advantages of multi-source measured data and model simulation to enhance monitoring accuracy and reliability.Changes in the water volume of the lower Lancang River have a thriving impact on downstream countries,but the water monitoring faces many challenges,such as the complexity of the basin and the impact of hydraulic engineering.In order to improve the accuracy and efficiency of water quantity monitoring,we proposed a nested multi-source data assimilation methodology coupled with the hydrodynamic model,which utilizes multi-source data such as manual measurement data,side-scan radar online monitoring data,and hydrodynamic numerical simulation data.Firstly,an accuracy improvement scheme of side-scan radar based on machine learning LASSO model was constructed by using artificial measured data.On this basis,a numerical simulation model of river hydrodynamics was constructed,and the parameters of hydrodynamic model were optimized by using the improved side-scan radar monitoring velocity,forming a nested fusion assimilation of multi-level and multi-source data.The aim was to improve the accuracy of water quantity simulation,extend the point observation data to the whole river,and expand the access range of the water quantity elements.The application in Yunjinghong Hydrological Station of Lancang River showed that:the accuracy improvement scheme based on machine learning LASSO model improved the accuracy of the side-scan radar monitoring system by 22.93%compared with the conventional method.The nested fusion assimilation mode of multi-level and multi-source data effectively improved the simulation accuracy of sectional flow,and the correlation coefficient in the verification period was 0.935.And the hydrological elements data such as water level,flow rate and velocity at any point in the modeled river channel were obtained.This study can provide a new methodology and technical support for the monitoring of water quantity of Lancang River.

关 键 词：多源数据 数据同化 水量监测 机器学习LASSO模型 水动力模型 侧扫雷达在线测流系统 澜沧江 

分 类 号：TV11[水利工程—水文学及水资源]