水动力学模型的实时组合校正方法研究  

Research on Real-time Combination Correction Method for Hydrodynamic Models

作　　者：马强[1] 陈璐 吴亚中 陈立正 贺阳 MA Qiang;CHEN Lu;WU Ya-zhong;CHEN Li-zheng;HE Yang(School of Civil and Hydraulic Engineering,Huazhong University of Science and Technology,Wuhan 430074,China;College of Water Conservancy and Civil Engineering,Xizang Agriculture and Animal Husbandry University,Nyingchi 860000,China)

机构地区：[1]华中科技大学土木与水利工程学院,湖北武汉430074 [2]西藏农牧学院水利土木工程学院,西藏林芝860000

出　　处：《水电能源科学》2025年第1期30-34,共5页Water Resources and Power

基　　金：国家重点研发计划(2023YFC3081000);西藏自治区自然科学基金项目(XZ202401ZR0044)。

摘　　要：受降雨分布不均、地形复杂多变及模型结构等因素限制,水动力学模型的水位模拟精度难以进一步提升。为此,基于一维水动力学模型运行结果,建立了涵盖KNN、SVR、LSTM、GRU等多种智能算法的实时校正模型,并将KNN校正法与WF算法串联耦合,提出了具有降噪功效的KNN-WF校正方法,在此基础上,为充分挖掘各校正算法在不同水位区间的校正优势,提出了一种基于M分位数的水位分级组合校正方法,以均方根误差、决定系数和峰值相对误差综合比较分析各校正方法的校正效果。汉江流域应用结果表明,KNN-WF校正法能有效避免KNN校正结果中存在的噪声和锯齿状峰值,校正后的水位过程与实测水位过程更加吻合,分级组合校正方法能实现不同水位区间的精细化校正从而提升整体校正效果。Due to factors such as uneven rainfall distribution,complex and varied terrain,and limitations of model structure,it is difficult to further improve the accuracy of water level simulation in hydrodynamic models.In this regard,a real-time correction model based on one-dimensional hydrodynamic model was developed using various intelligent algo-rithms,including K-Nearest Neighbors,Support Vector Regression,Long Short-Term Memory,and Gated Recurrent Unit.Furthermore,the KNN correction method was combined with the Wavelet Filtering algorithm to propose a KNN-WF correction method,which effectively reduces noise and jagged peaks in the KNN correction results.In addition,a graded combination correction method based on the M-Quantile method was proposed to fully exploit the correction ad-vantages of each algorithm in different water level intervals.The correction effects of various correction methods were compared and analyzed comprehensively using Root Mean Square Error,Determination Coefficient,and Peak Relative Er-ror.The results obtained from application in the Hanjiang River Basin demonstrate that the KNN-WF correction method can effectively mitigate the noise and jagged peaks present in the KNN correction results,resulting in a water level process that aligns better with the observed water level.The segmented combination correction method achieves fine-grained correction in different water level intervals,thereby enhancing the overall correction effect.

关 键 词：水动力学模型 终端校正 分级组合校正 汉江流域 

分 类 号：TV131.2[水利工程—水力学及河流动力学]