Core Concepts
Multi-resolution dynamic mode decomposition (MR-DMD) can effectively identify dominant oscillation modes and reconstruct measured signals in power electronic converter systems, even with transient time behaviors in the dataset.
Abstract
The paper presents a data-driven diagnostic tool called multi-resolution dynamic mode decomposition (MR-DMD) for assessing the stability of power electronic converter systems. MR-DMD combines the conventional dynamic mode decomposition (DMD) algorithm with multi-resolution analysis from wavelet theory to hierarchically decompose datasets into multiple time bins and apply DMD recursively.
Key highlights:
- MR-DMD can identify dominant oscillation modes and eigenvalues at different frequency ranges and time scales, addressing the limitations of conventional DMD in handling datasets with transient time behaviors.
- The paper derives the selection criteria for three crucial parameters of MR-DMD - the fixed subsample number of each time bin, the termination level, and the screening threshold of slow modes. This elucidates how the algorithm can extract eigenvalues within different frequency ranges.
- The performance of MR-DMD is demonstrated using an experimental dataset from a single-phase converter platform experiencing low-frequency oscillations. Compared to conventional DMD, MR-DMD can accurately reconstruct the measured signals and identify the dominant oscillation modes, even with missing data in the dataset.
- The proposed methodology is generic and can be applied to a wide range of power electronic converter systems for stability assessment.
Stats
The experimental system has the following main parameters:
Grid phase voltage (RMS): 110 V
Grid-side inductance: 8 mH
Converter-side inductance: 4 mH
Filter capacitance: 10 μF
DC link voltage: 170 V
DC link load resistance: 460 Ω
DC link support capacitance: 800 μF
Switching frequency: 10 kHz
Quotes
"By combining dynamic mode decomposition (DMD) with the multi-resolution analysis used in wavelet theory, dynamic modes and eigenvalues can be identified at different decomposition levels and time scales with the MR-DMD algorithm, thereby allowing for handling datasets with transient time behaviors, which is not achievable using conventional DMD."
"The selection criteria for important parameters in MR-DMD are clearly defined through derivation, elucidating the reason for enabling it to extract eigenvalues within different frequency ranges."