| 王家彪,赵建世,雷晓辉,王浩,魏隽煜,廖卫红.基于EnKF的无实测资料区间支流反分析[J].水利学报,2019,50(10):1189-1199 |
| 基于EnKF的无实测资料区间支流反分析 |
| Inverse simulation of ungauged branch inflow based on the EnKF |
| 投稿时间：2019-06-18 |
| DOI：10.13243/j.cnki.slxb.20190442 |
| 中文关键词: 反问题 数据同化 反向水流计算 滞时矩阵 EnKF |
| 英文关键词: inverse problem data assimilation reverse flow routing lag-time matrix EnKF |
| 基金项目:国家自然科学基金项目（91747208；51579129） |
| 作者 | 单位 | E-mail | | 王家彪 | 清华大学, 水沙科学与水利水电工程国家重点实验室, 北京 100084 | | | 赵建世 | 清华大学, 水沙科学与水利水电工程国家重点实验室, 北京 100084 | | | 雷晓辉 | 中国水利水电科学研究院, 北京 100038 | lxh@iwhr.com | | 王浩 | 清华大学, 水沙科学与水利水电工程国家重点实验室, 北京 100084
中国水利水电科学研究院, 北京 100038 | | | 魏隽煜 | 清华大学, 水沙科学与水利水电工程国家重点实验室, 北京 100084 | | | 廖卫红 | 中国水利水电科学研究院, 北京 100038 | |
|
| 摘要点击次数: 1255 |
| 全文下载次数: 752 |
| 中文摘要: |
| 区间入流误差是河道洪水演算不确定性来源之一。为此，一部分研究基于水动力模型和数据同化方法对区间总入流误差进行动态修正，但无法推算出某条支流的单独入流过程；另一部分研究通过从下游往上游反算水流的方式推算区间支流入流，但反算结果稳定性差，对边界条件误差敏感，推算的入流过程误差较大。针对上述问题，本文提出了基于集合卡尔曼滤波（EnKF）的区间支流反分析方法。方法由一维河道水动力模型正、反向水流演算初步估算支流入流，并构建监测断面滞时矩阵，计算水流扰动传播时间，从而确定用于支流入流校正的流量监测值。当EnKF校正的结果仍然存在较大误差时，可再次运用EnKF对首次校正结果进一步校正。将该方法应用于理想案例和西江实例，推算的支流入流过程与实测过程十分接近，入流结果R2和NSE皆在0.99以上，相对RMSE也小于0.05。结果表明，本文提出方法可准确计算出无实测资料的区间支流入流过程，研究结果对于提高河道洪水演算精度具有重要意义。 |
| 英文摘要: |
| The lateral inflow error is one the main uncertainty sources of river flood routing. Parts of previous methods for inverse simulation of lateral inflow are not able to determine the separate inflow of a specific branch,because they commonly calculate the total lateral inflow based on the principle of water balance combined with hydrodynamic models and data assimilation. In other parts of previous methods, the reverse flood routing from downstream to upstream has ever been attempted for the inverse analysis of branch inflow. However, the results from reverse flood routing is sensitive to the boundary conditions, and significant uncertainties lead to large errors in the inverse simulation results. In terms of above-mentioned issues, an EnKF-based method is proposed for the inverse simulation of ungauged branch inflow. The branch inflow is firstly calculated via the forward and reverse flood routing with one-dimensional hydrodynamic model. Then, the lag-time matrix used for EnKF is determined by the propagation time of perturbation from branch to the gauging station. When large errors are obtained in the first-time calibration results, the EnKF should be reused with the first results as the background. The EnKF-based method is applied to an ideal case and a real-world case in Xijiang River,and the inverse simulation results are pretty good. In both cases,the assessment indicators R2 and NSE are above 0.99,and the relative RMSE is less than 0.05. The case studies demonstrate the efficiency and effectiveness of our proposed method in inverse simulation of ungauged branch inflow,and the research is helpful to improve the simulation accuracy of flood routing. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|