Estimate of Koopman modes and eigenvalues with Kalman Filter

Abstract

Dynamic mode decomposition (DMD) is a data-driven method of extracting spatial-temporal coherent modes from complex systems and providing an equation-free architecture to model and predict systems. However, in practical applications, the accuracy of DMD can be limited in extracting dynamical features due to sensor noise in measurements. We develop an adaptive method to constantly update dynamic modes and eigenvalues from noisy measurements arising from discrete systems. Our method is based on the Ensemble Kalman filter owing to its capability of handling time-varying systems and nonlinear observables. Our method can be extended to non-autonomous dynamical systems, accurately recovering short-time eigenvalue-eigenvector pairs and observables. Theoretical analysis shows that the estimation is accurate in long term data misfit. We demonstrate the method on both autonomous and non-autonomous dynamical systems to show its effectiveness.

Figures (12)

• Figure 1: Flowchart for determining the numeber of delay coordinates.
• Figure 1: Reconstruction of observable functions.
• Figure 2: Reconstruction and prediction of observable functions.
• Figure 3: Five coefficients in the Fourier domain and spatial structures of the dynamical system.
• Figure 4: True DMD modes, modes estimated by compressed DMD and EnKF-DMD.

Theorems & Definitions (11)

• Remark 4.2
• Theorem 4.3
• Theorem 4.5
• Proof 1: Proof for \ref{['thm:Auto_Converge']}
• Lemma A.1
• Proof 2
• Lemma A.2
• Proof 3
• Lemma A.3
• Proof 4