pith. sign in

arxiv: 2502.09592 · v2 · pith:4UM7ULF2new · submitted 2025-02-13 · 📡 eess.SY · cs.SY

A Data-Driven Method for Microgrid System Identification: Physically Consistent Sparse Identification of Nonlinear Dynamics

Mohan Du , Xiaozhe Wang This is my paper

Pith reviewed 2026-05-23 03:26 UTC · model grok-4.3

classification 📡 eess.SY cs.SY
keywords microgrid system identificationsparse identification of nonlinear dynamicsPMU datafrequency trajectory predictionphysically consistent modelingdata-driven dynamicspower system stabilitynonlinear system identification
0
0 comments X

The pith

PC-SINDy extracts accurate microgrid dynamic models from PMU data that predict frequency trajectories under large unseen disturbances.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper introduces the Physically Consistent Sparse Identification of Nonlinear Dynamics (PC-SINDy) method for microgrid system identification. It leverages an analytically derived library of candidate functions to perform sparse regression solely on phasor measurement unit data. The goal is to recover models that accurately forecast system behavior without detailed prior knowledge of topology or parameters. Simulations on a 4-bus system demonstrate reliable predictions even with noisy low-sampled data and for disturbances absent from the training set.

Core claim

By using an analytically derived library of candidate functions within the sparse identification of nonlinear dynamics framework, PC-SINDy recovers dynamic models of microgrids from PMU data alone that reliably and accurately predict frequency trajectories under large disturbances, including those not encountered during identification, even with noisy and low-sampled measurements.

What carries the argument

Physically Consistent Sparse Identification of Nonlinear Dynamics (PC-SINDy) using an analytically derived library of candidate functions for sparse regression on PMU data.

If this is right

  • Microgrid models can be identified without explicit knowledge of system parameters or topology.
  • The recovered models support design of robust control strategies for microgrid stability.
  • Predictions remain accurate when input data are noisy and sampled at low rates.
  • Models generalize to large disturbances outside the training dataset.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same library-based sparse regression approach could be tested on microgrids with higher numbers of distributed energy resources.
  • Streaming PMU data could enable periodic re-identification to track slow changes in topology or operating modes.
  • Direct comparison of prediction error against purely physics-based models on the same 4-bus system would quantify the data-driven advantage.

Load-bearing premise

The analytically derived library of candidate functions contains every term required to represent the true microgrid dynamics so sparse regression can recover the correct model.

What would settle it

A test case in which the PC-SINDy model recovered from noisy PMU data on the 4-bus system produces frequency trajectory predictions that diverge from measured responses under a large disturbance absent from training data would falsify the claim.

Figures

Figures reproduced from arXiv: 2502.09592 by Mohan Du, Xiaozhe Wang.

Figure 1
Figure 1. Figure 1: (a) Droop control of GFM converters and (b) three-phase PLL [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Flowchart of the PC-SINDy-based Identification method for the true physical system. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison between analytical Ξ and SINDy-estimated Ξˆ using low￾quality measurements [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: The 4-bus MG. A. System Identification by PC-SINDy Using Noisy Measure￾ments with Low-Sampling Rate In Step 1, continuous random inputs u(t) are generated with a duration of tM = 10 s, ranging within ±0.01 p.u., to excite the physical system. Next, in Step 2, PMU measurements are collected with noise levels and sampling rates compliant with IEEE C37.118.1-2011 [17], which are then used to construct X˙ and … view at source ↗
Figure 5
Figure 5. Figure 5: Prediction of the GFM frequency f1 and the GFL frequency f2 following (17) with the proposed analytical library and intuitive library [2]. and sinusoids of X and U [2]. The corresponding coefficient matrix is extracted from PMU measurements and then used for predictions on the validation set [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
read the original abstract

Microgrids (MGs) play a crucial role in utilizing distributed energy resources (DERs) like solar and wind power, enhancing the sustainability and flexibility of modern power systems. However, the inherent variability in MG topology, power flow, and DER operating modes poses significant challenges to the accurate system identification of MGs, which is crucial for designing robust control strategies and ensuring MG stability. This paper proposes a Physically Consistent Sparse Identification of Nonlinear Dynamics (PC-SINDy) method for accurate MG system identification. By leveraging an analytically derived library of candidate functions, PC-SINDy extracts accurate dynamic models using only phasor measurement unit (PMU) data. Simulations on a 4-bus system demonstrate that PC-SINDy can reliably and accurately predict frequency trajectories under large disturbances, including scenarios not encountered during the identification/training phase, even when using noisy, low-sampled PMU data.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper proposes Physically Consistent Sparse Identification of Nonlinear Dynamics (PC-SINDy) for microgrid system identification. It constructs an analytically derived library of candidate functions from swing equations and DER models, then applies sparse regression to PMU data to recover dynamic models. Simulations on a 4-bus system are used to claim that the resulting models accurately predict frequency trajectories under large disturbances (including out-of-sample cases) even with noisy, low-sampled PMU measurements.

Significance. If the library-completeness and sparse-recovery assumptions hold, the approach would offer a practical route to data-driven MG modeling that generalizes beyond training regimes without requiring full topology or parameter knowledge, which is valuable for control design in variable-topology systems with high DER penetration.

major comments (3)
  1. [Section 3 (Library Construction) and Section 5 (Numerical Results)] The load-bearing assumption is that the analytically derived candidate library is exhaustive for the true dynamics under large disturbances not seen in training. No section provides an explicit completeness argument or sensitivity test showing that omitted terms cannot produce the reported out-of-sample accuracy; if any relevant term is missing, the sparse regression cannot recover the correct model.
  2. [Section 5 (Numerical Results)] Section 5 reports that PC-SINDy 'reliably and accurately' predicts frequency trajectories on the 4-bus system, yet supplies no quantitative error metrics (RMSE, NRMSE, or maximum deviation) nor ablation on noise level and sampling rate; without these, the generalization claim cannot be evaluated against the stated conditions of noisy, low-sampled PMU data.
  3. [Section 4 (PC-SINDy Algorithm)] The sparse-regression step is asserted to select the correct terms without spurious or omitted dynamics, but the manuscript does not report the regularization parameter, threshold values, or cross-validation procedure used to guarantee uniqueness under realistic noise; this directly affects whether the recovered model is physically consistent or merely data-fit.
minor comments (2)
  1. [Figures 3-5] Figure captions should explicitly state the noise variance, sampling frequency, and disturbance magnitudes used in each panel so readers can reproduce the claimed robustness.
  2. [Section 2] The notation for the physically consistent constraints (e.g., power-balance or passivity conditions) is introduced without a compact equation reference; adding a single displayed equation would improve clarity.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their thorough and constructive review. We address each major comment below and will revise the manuscript accordingly to improve rigor, transparency, and evaluability of the claims.

read point-by-point responses

  1. Referee: [Section 3 (Library Construction) and Section 5 (Numerical Results)] The load-bearing assumption is that the analytically derived candidate library is exhaustive for the true dynamics under large disturbances not seen in training. No section provides an explicit completeness argument or sensitivity test showing that omitted terms cannot produce the reported out-of-sample accuracy; if any relevant term is missing, the sparse regression cannot recover the correct model.

    Authors: The candidate library is constructed by direct analytical derivation from the swing equations and standard DER dynamic models, which are the governing nonlinear equations expected to hold for large disturbances. We agree that an explicit completeness argument and sensitivity discussion are absent and would strengthen the paper. In revision we will expand Section 3 with a detailed derivation of each term and a discussion of why the library is expected to be complete under the modeled physics, plus a brief sensitivity analysis in Section 5. revision: yes

  2. Referee: [Section 5 (Numerical Results)] Section 5 reports that PC-SINDy 'reliably and accurately' predicts frequency trajectories on the 4-bus system, yet supplies no quantitative error metrics (RMSE, NRMSE, or maximum deviation) nor ablation on noise level and sampling rate; without these, the generalization claim cannot be evaluated against the stated conditions of noisy, low-sampled PMU data.

    Authors: We agree that quantitative metrics and ablations are necessary to substantiate the generalization claims under noisy, low-sampled conditions. The current Section 5 uses qualitative descriptions. In the revised manuscript we will add RMSE, NRMSE, and maximum deviation values for the reported trajectories together with results showing performance across multiple noise levels and sampling rates. revision: yes

  3. Referee: [Section 4 (PC-SINDy Algorithm)] The sparse-regression step is asserted to select the correct terms without spurious or omitted dynamics, but the manuscript does not report the regularization parameter, threshold values, or cross-validation procedure used to guarantee uniqueness under realistic noise; this directly affects whether the recovered model is physically consistent or merely data-fit.

    Authors: The referee is correct that specific hyperparameter values and the selection procedure are required for reproducibility and to evaluate robustness under noise. Although Section 4 describes the overall sparse-regression approach, these details are not provided. In revision we will report the regularization parameter, threshold values, and cross-validation procedure used, allowing readers to assess uniqueness and physical consistency. revision: yes

Circularity Check

0 steps flagged

No significant circularity; library is external physics input, predictions are out-of-sample tests

full rationale

The derivation uses an analytically derived candidate library (presumably from swing equations/DER models) as input to sparse regression on PMU data, then evaluates frequency predictions on large disturbances outside the training set. This structure does not reduce the reported predictions to fitted quantities by construction, nor does it rely on self-citation chains or self-definitional loops. The library-completeness assumption is a modeling choice, not a circular reduction. Score remains low (1) as a minor allowance for any unexamined self-citation in the full text.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no information on free parameters, axioms, or invented entities.

pith-pipeline@v0.9.0 · 5685 in / 964 out tokens · 30856 ms · 2026-05-23T03:26:46.619590+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

21 extracted references · 21 canonical work pages

  1. [1]

    On data-driven modeling and control in modern power grids stability: Survey and perspective,

    X. Gong, X. Wang, and B. Cao, “On data-driven modeling and control in modern power grids stability: Survey and perspective,” Applied Energy, vol. 350, p. 121740, Nov. 2023

    work page 2023

  2. [2]

    Data-Driven Modeling of Microgrid Transient Dynam- ics through Modularized Sparse Identification,

    A. Nandakumar, Y . Li, H. Zheng, J. Zhao, D. Zhao, Y . Zhang, T. Hong, and B. Chen, “Data-Driven Modeling of Microgrid Transient Dynam- ics through Modularized Sparse Identification,” IEEE Transactions on Sustainable Energy, pp. 1–14, 2023

    work page 2023

  3. [3]

    Fusion of Microgrid Control With Model-Free Reinforcement Learning: Review and Vision,

    B. She, F. Li, H. Cui, J. Zhang, and R. Bo, “Fusion of Microgrid Control With Model-Free Reinforcement Learning: Review and Vision,” IEEE Transactions on Smart Grid , vol. 14, no. 4, pp. 3232–3245, Jul. 2023

    work page 2023

  4. [4]

    Interactive Control of Coupled Microgrids for Guaranteed System-Wide Small Signal Stability,

    Y . Zhang, L. Xie, and Q. Ding, “Interactive Control of Coupled Microgrids for Guaranteed System-Wide Small Signal Stability,” IEEE Transactions on Smart Grid , vol. 7, no. 2, pp. 1088–1096, Mar. 2016

    work page 2016

  5. [5]

    Distributed Cooperative Secondary Control of Microgrids Using Feedback Lin- earization,

    A. Bidram, A. Davoudi, F. L. Lewis, and J. M. Guerrero, “Distributed Cooperative Secondary Control of Microgrids Using Feedback Lin- earization,” IEEE Transactions on Power Systems , vol. 28, no. 3, pp. 3462–3470, Aug. 2013

    work page 2013

  6. [6]

    A novel Koopman-inspired method for the secondary control of microgrids with grid-forming and grid-following sources,

    X. Gong and X. Wang, “A novel Koopman-inspired method for the secondary control of microgrids with grid-forming and grid-following sources,” Applied Energy, vol. 333, p. 120631, Mar. 2023

    work page 2023

  7. [7]

    Hierarchical Control of Parallel AC-DC Converter Interfaces for Hybrid Microgrids,

    X. Lu, J. M. Guerrero, K. Sun, J. C. Vasquez, R. Teodorescu, and L. Huang, “Hierarchical Control of Parallel AC-DC Converter Interfaces for Hybrid Microgrids,” IEEE Transactions on Smart Grid, vol. 5, no. 2, pp. 683–692, Mar. 2014

    work page 2014

  8. [8]

    Consensus-Based Delay- Tolerant Distributed Secondary Control Strategy for Droop Controlled AC Microgrids,

    S. Ullah, L. Khan, I. Sami, and N. Ullah, “Consensus-Based Delay- Tolerant Distributed Secondary Control Strategy for Droop Controlled AC Microgrids,” IEEE Access, vol. 9, pp. 6033–6049, 2021

    work page 2021

  9. [9]

    Data-Driven LPV Controller Design for Islanded Microgrids *,

    S. S. Madani and A. Karimi, “Data-Driven LPV Controller Design for Islanded Microgrids *,” IFAC-PapersOnLine, vol. 54, no. 7, pp. 433– 438, Jan. 2021

    work page 2021

  10. [10]

    Recurrent neural networks for phasor detection and adaptive identifi- cation in power system control and protection,

    I. Kamwa, R. Grondin, V . Sood, C. Gagnon, V . T. Nguyen, and J. Mereb, “Recurrent neural networks for phasor detection and adaptive identifi- cation in power system control and protection,” IEEE Transactions on Instrumentation and Measurement , vol. 45, no. 2, pp. 657–664, Apr. 1996

    work page 1996

  11. [11]

    Intelligent Multi-Microgrid Energy Management Based on Deep Neural Network and Model-Free Reinforcement Learning,

    Y . Du and F. Li, “Intelligent Multi-Microgrid Energy Management Based on Deep Neural Network and Model-Free Reinforcement Learning,” IEEE Transactions on Smart Grid , vol. 11, no. 2, pp. 1066–1076, Mar. 2020

    work page 2020

  12. [12]

    Chapter 12: Neuroscience,

    J. N. Kutz, S. L. Brunton, B. W. Brunton, and J. L. Proctor, “Chapter 12: Neuroscience,” inDynamic Mode Decomposition, ser. Other Titles in Applied Mathematics. Society for Industrial and Applied Mathematics, Nov. 2016, pp. 185–193

    work page 2016

  13. [13]

    An Online Data-Driven Method to Locate Forced Oscillation Sources From Power Plants Based on Sparse Identification of Nonlinear Dynamics (SINDy),

    Y . Cai, X. Wang, G. Jo ´os, and I. Kamwa, “An Online Data-Driven Method to Locate Forced Oscillation Sources From Power Plants Based on Sparse Identification of Nonlinear Dynamics (SINDy),” IEEE Transactions on Power Systems , vol. 38, no. 3, pp. 2085–2099, May 2023

    work page 2085

  14. [14]

    Physics-informed learning of governing equations from scarce data,

    Z. Chen, Y . Liu, and H. Sun, “Physics-informed learning of governing equations from scarce data,” Nature Communications, vol. 12, no. 1, p. 6136, Oct. 2021

    work page 2021

  15. [15]

    Analytical Large-Signal Modeling of Inverter-Based Microgrids With Koopman Operator Theory for Au- tonomous Control,

    Z. Ma, Z. Wang, and R. Cheng, “Analytical Large-Signal Modeling of Inverter-Based Microgrids With Koopman Operator Theory for Au- tonomous Control,” IEEE Transactions on Smart Grid , vol. 15, no. 2, pp. 1376–1387, Mar. 2024

    work page 2024

  16. [16]

    An Online Data-Driven Method for Microgrid Secondary V oltage and Frequency Control With Ensemble Koopman Modeling,

    X. Gong, X. Wang, and G. Joos, “An Online Data-Driven Method for Microgrid Secondary V oltage and Frequency Control With Ensemble Koopman Modeling,” IEEE Transactions on Smart Grid , vol. 14, no. 1, pp. 68–81, Jan. 2023

    work page 2023

  17. [17]

    IEEE Standard for Synchrophasor Measurements for Power Systems,

    “IEEE Standard for Synchrophasor Measurements for Power Systems,” IEEE Std C37.118.1-2011 (Revision of IEEE Std C37.118-2005) , pp. 1–61, Dec. 2011

    work page 2011

  18. [18]

    Chapter 3: Koopman Analysis,

    J. N. Kutz, S. L. Brunton, B. W. Brunton, and J. L. Proctor, “Chapter 3: Koopman Analysis,” in Dynamic Mode Decomposition , ser. Other Titles in Applied Mathematics. Society for Industrial and Applied Mathematics, Nov. 2016, pp. 39–53

    work page 2016

  19. [19]

    Grid-imposed frequency VSC system: Control in dq-frame,

    A. Yazdani and R. Iravani, “Grid-imposed frequency VSC system: Control in dq-frame,” in Voltage-Sourced Converters in Power Systems: Modeling, Control, and Applications , 2010, pp. 204–244

    work page 2010

  20. [20]

    Phase-Locked Loops - MATLAB & Simulink,

    “Phase-Locked Loops - MATLAB & Simulink,” https://www.mathworks.com/help/msblks/phase-locked-loop.html

    work page

  21. [21]

    Discovering governing equations from data by sparse identification of nonlinear dynamical systems,

    S. L. Brunton, J. L. Proctor, and J. N. Kutz, “Discovering governing equations from data by sparse identification of nonlinear dynamical systems,” Proceedings of the National Academy of Sciences , vol. 113, no. 15, pp. 3932–3937, Apr. 2016

    work page 2016