Performance assessment of the maximum likelihood ensemble filter and the ensemble Kalman filters for nonlinear problems

Performance assessment of the maximum likelihood ensemble filter and the ensemble Kalman filters for nonlinear problems
复制标题

针对非线性问题的最大似然集成滤波器和集成卡尔曼滤波器的性能评估

DOI:
10.1007/s40687-022-00359-7
复制
发表时间:
2022
影响因子:
1.2
通讯作者:
Gao, Xinfeng
Gao, Xinfeng
中科院分区:
数学3区
文献类型:
--
作者:
Wang, Yijun;Zupanski, Milija;Tu, Xuemin;Gao, Xinfeng

文献摘要

参考文献

相似文献

本研究对三种集合数据同化(DA)方法(包括最大似然集合滤波器(MLEF)、集合卡尔曼滤波器(EnKF)和迭代 EnKF(IEnKF))在非线性问题的求解精度和计算效率方面的性能比较进行了深入研究。首先测试对流扩散反应(CDR)问题,然后求解混沌 Lorenz 96 模型。线性和非线性观测算子都被考虑。研究表明,MLEF 始终能够比其他两种方法产生更准确、更高效的解决方案,并提供有关两种状态及其不确定性的更多信息。 IEnKF 和 MLEF 用于使用非线性观测算子估计初始条件下的模型参数和不确定性。同化性能是根据质量指标进行评估的,例如平方真实误差、误差协方差矩阵的迹和均方根(RMS)误差。基于这些 DA 性能评估,MLEF 表现出更好的收敛性和更高的准确性。 CDR 问题的结果表明,与 EnKF 系列相比,MLEF 在模型参数估计和求解精度方面有显着改进。这项研究提供了支持在解决大型非线性问题时选择 MLEF 的证据。
This study presents a thorough investigation of the performance comparison of three ensemble data assimilation (DA) methods, including the maximum likelihood ensemble filter (MLEF), the ensemble Kalman filter (EnKF), and the iterative EnKF (IEnKF), with respect to solution accuracy and computational efficiency for nonlinear problems. The convection–diffusion–reaction (CDR) problem is first tested, and then, the chaotic Lorenz 96 model is solved. Both linear and nonlinear observation operators are considered. The study demonstrates that MLEF consistently produces more accurate and efficient solution than the other two methods and provides more information on both states and their uncertainties. The IEnKF and MLEF are used to estimate model parameters and uncertainty in initial conditions using a nonlinear observation operator. The assimilation performance is assessed based on the quality metrics, such as the squared true error, the trace of the error covariance matrix, and the root-mean-square (RMS) error. Based on these DA performance assessments, MLEF demonstrates better convergence and higher accuracy. Results of the CDR problem show significant improvements in the estimate of model parameters and the solution accuracy by MLEF compared to the EnKF family. This study provides evidence supporting the choice of MLEF when solving large nonlinear problems.
用于燃烧应用的改进集成卡尔曼滤波器算法的特性
DOI: 10.2514/6.2016-3484
发表时间: 2016
期刊: Comput. Math. Appl.
影响因子: --
作者:
Xinfeng Gao;Yijun Wang;Nathaniel Overton;M. Zupanski;Xuemin Tu
通讯作者: Xuemin Tu
利用数据同化对钝体稳定预混火焰进行 CFD 建模
DOI: 10.2514/6.2020-0352
发表时间: 2020
期刊: AIAA meeting papers on disc
影响因子: --
作者:
Wang, Y.;Walters, S.;Overton, N.;Guzik, S. M.;Gao, X.
通讯作者: Gao, X.
通过数据同化提高湍流可预测性的有限极限
DOI: 10.2514/6.2020-0350
发表时间: 2020
影响因子: 8.9
作者:
C. Hurst;Xinfeng Gao
通讯作者: Xinfeng Gao
DOI: 10.2307/2346308
发表时间: 1973
影响因子: 1.6
作者:
H. Neave
通讯作者: H. Neave
用于计算火焰和流体动力学的最大似然系综滤波器
DOI: 10.1093/imamat/hxab010
发表时间: 2021
影响因子: 1.2
作者:
Yijun Wang;S. Guzik;M. Zupanski;Xinfeng Gao
通讯作者: Xinfeng Gao