Deep learning-enhanced ensemble-based data assimilation for high-dimensional nonlinear dynamical systems

Deep learning-enhanced ensemble-based data assimilation for high-dimensional nonlinear dynamical systems
复制标题

高维非线性动力系统的深度学习增强型基于集成的数据同化

DOI:
10.1016/j.jcp.2023.111918
复制
发表时间:
2023
影响因子:
4.1
通讯作者:
Hassanzadeh, Pedram
Hassanzadeh, Pedram
中科院分区:
物理与天体物理2区
文献类型:
--
作者:
Chattopadhyay, Ashesh;Nabizadeh, Ebrahim;Bach, Eviatar;Hassanzadeh, Pedram

文献摘要

参考文献

被引文献

相似文献

数据同化(DA)是科学和工程中许多预测模型的关键组成部分。 DA 允许人们使用系统的不完美动态模型和系统中可用的噪声/稀疏观测来估计更好的初始条件。集成卡尔曼滤波器 (EnKF) 是一种 DA 算法,广泛应用于涉及高维非线性动力系统的应用中。然而,EnKF 需要使用系统的动态模型发展大型预测集合。这通常在计算上变得困难,特别是当系统的状态数量非常大时,例如天气预报。对于小型集成,EnKF 算法中估计的背景误差协方差矩阵会受到采样误差的影响,从而导致对分析状态(下一个预测周期的初始条件)的错误估计。在这项工作中,我们提出了混合系综卡尔曼滤波器(H-EnKF),将其应用于两层准地转湍流作为测试用例。该框架利用预先训练的基于深度学习的数据驱动代理,以低廉的成本生成和演化大型数据驱动的状态集合,以较小的采样误差准确计算背景误差协方差矩阵。 H-EnKF 框架优于仅使用动态模型或仅使用数据驱动代理的 EnKF,并且无需任何临时本地化策略即可估计更好的初始条件。 H-EnKF 可以扩展到任何基于集成的 DA 算法,例如粒子滤波器,目前这些算法对于高维系统来说过于昂贵。
Data assimilation (DA) is a key component of many forecasting models in science and engineering. DA allows one to estimate better initial conditions using an imperfect dynamical model of the system and noisy/sparse observations available from the system. Ensemble Kalman filter (EnKF) is a DA algorithm that is widely used in applications involving high-dimensional nonlinear dynamical systems. However, EnKF requires evolving large ensembles of forecasts using the dynamical model of the system. This often becomes computationally intractable, especially when the number of states of the system is very large, e.g., for weather prediction. With small ensembles, the estimated background error covariance matrix in the EnKF algorithm suffers from sampling error, leading to an erroneous estimate of the analysis state (initial condition for the next forecast cycle). In this work, we propose hybrid ensemble Kalman filter (H-EnKF), which is applied to a two-layer quasi-geostrophic turbulent flow as a test case. This framework utilizes a pre-trained deep learning-based data-driven surrogate that inexpensively generates and evolves a large data-driven ensemble of the states to accurately compute the background error covariance matrix with smaller sampling errors. The H-EnKF framework outperforms EnKF with only dynamical model or only the data-driven surrogate, and estimates a better initial condition without the need for any ad-hoc localization strategies. H-EnKF can be extended to any ensemble-based DA algorithm, e.g., particle filters, which are currently too expensive to use for high-dimensional systems.
DOI: 10.1175/mwr-d-15-0388.1
发表时间: 2016-11
影响因子: 3.2
作者:
K. Kondo;T. Miyoshi
通讯作者: K. Kondo;T. Miyoshi
DOI: 10.1016/j.combustflame.2019.04.023
发表时间: 2018-03
影响因子: 4.4
作者:
J. Bell;M. Day;J. Goodman;R. Grout;M. Morzfeld
通讯作者: J. Bell;M. Day;J. Goodman;R. Grout;M. Morzfeld
用于数据同化的滤波器和平滑器中的高斯近似
DOI: 10.1080/16000870.2019.1600344
发表时间: 2019
期刊: Tellus A: Dynamic Meteorology and Oceanography
影响因子: --
作者:
Morzfeld, Matthias;Hodyss, Daniel
通讯作者: Hodyss, Daniel
使用异方差贝叶斯神经网络集成进行降阶火焰模型的数据同化
DOI: 10.1007/978-3-030-77977-1_33
发表时间: 2021
期刊: Frontiers Appl. Math. Stat.
影响因子: --
作者:
Maximilian L. Croci;Ushnish Sengupta;M. Juniper
通讯作者: M. Juniper
DOI: 10.1016/j.jcp.2018.10.042
发表时间: 2019-02-15
影响因子: 4.1
作者:
Arcucci, Rossella;Mottet, Laetitia;Guo, Yi-Ke
通讯作者: Guo, Yi-Ke