pith. sign in

arxiv: 2604.12067 · v2 · submitted 2026-04-13 · 💻 cs.IT · cs.SY· eess.SY· math.IT

Vectorized Gaussian Belief Propagation for Near Real-Time Fully-Distributed PMU-Based State Estimation

Mirsad Cosovic , Armin Teskeredzic , Antonello Monti , Dejan Vukobratovic This is my paper

Pith reviewed 2026-05-10 14:53 UTC · model grok-4.3

classification 💻 cs.IT cs.SYeess.SYmath.IT
keywords state estimationbelief propagationfactor graphsdistributed algorithmsphasor measurement unitspower systemsGaussian modelsreal-time monitoring
0
0 comments X

The pith

A vectorized Gaussian belief propagation approach delivers near real-time distributed state estimation for large power systems with phasor measurements.

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

This paper develops a framework using Gaussian belief propagation on factor graphs to perform state estimation in electric power systems based on phasor measurement units. The method supports two formulations: one that jointly models related variables and another that fuses measurements to simplify the graph structure. By enabling computations to occur locally at each bus without a central coordinator, it addresses the need for scalable and fast monitoring in complex grids. Tests demonstrate that these algorithms converge quickly with high accuracy on systems as large as 13,659 buses, with the fusion approach achieving millisecond-level iteration times.

Core claim

The paper establishes that the proposed vectorized Gaussian belief propagation algorithms, running in a fully distributed manner at the bus level, achieve fast convergence and high estimation accuracy for phasor measurement unit-based state estimation, often within a few iterations. This is validated through numerical results on power systems ranging from 60 to 13,659 buses, where the fusion-based formulation attains single-digit millisecond iteration times on the largest case.

What carries the argument

The vectorized Gaussian belief propagation framework over factor graphs, with multivariate and fusion-based formulations that allow local message passing for global state estimates.

If this is right

  • The method operates fully distributed at the bus level without central coordination.
  • It achieves high estimation accuracy often within a few iterations.
  • The fusion-based formulation reduces complexity by combining related measurements.
  • Performance scales to very large systems with millisecond iteration times.

Where Pith is reading between the lines

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

  • Such distributed estimation could support real-time control applications in smart grids by providing timely data without communication bottlenecks.
  • Extending the framework to handle dynamic state estimation or incorporating renewable energy variability might be a natural next step.
  • The factor graph approach could apply to other sensor networks beyond power systems for distributed inference tasks.

Load-bearing premise

The power system and its measurements can be accurately represented by a factor graph with Gaussian probability distributions, allowing local message passing to preserve overall estimation accuracy without central coordination or modeling errors.

What would settle it

Running the algorithm on a large power grid model and comparing its estimates to a centralized state estimator; if the distributed version shows significantly lower accuracy or fails to converge on certain configurations, the effectiveness claim would be disproven.

Figures

Figures reproduced from arXiv: 2604.12067 by Antonello Monti, Armin Teskeredzic, Dejan Vukobratovic, Mirsad Cosovic.

Figure 1
Figure 1. Figure 1: Transformation of the bus/branch model and measurem [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Message µfk→xi (xi) from factor node fk to variable node xi. The message µfk→xi (xi) is obtained by multiplying the incoming message µxj→fk (xj ) with the local likelihood func￾tion p(zfk |xi , xj ) defined in (5), and marginalising over the variable xj : µfk→xi (xi) = Z p(zfk |xi , xj ) µxj→fk (xj ) dxj . (11) Once the incoming message µxj→fk (xj ) is available, char￾acterised by the mean zxj→fk and preci… view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of the initialization procedure for th [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Messages µfk→xi (xi) from pairwise factor nodes to variable nodes. Consider the current phasor measurement associated with factor node f4, modeled according to (5). The measurement value is zf4 = [−0.28, −0.61]T , with precision matrix Λf4 = 105 · diag(0.27, 1.12), and coefficient matrices: Hf4,x1 =  2.0 4.0 −4.0 2.0  , Hf4,x3 =  −2.0 −4.0 4.0 −2.0  . (15) For illustration, consider the message from fa… view at source ↗
Figure 6
Figure 6. Figure 6: Messages µxi→fk (xi) from variable nodes to pairwise factor nodes. Consider the message from variable node x3 to factor node f6. The incoming message, characterized by the mean zf4→x3 and precision matrix Λf4→x3 , is obtained from the previous step, in which messages from factor nodes to variable nodes are computed. Since only a single incoming message is present, the message to factor node f6 is equal to … view at source ↗
Figure 5
Figure 5. Figure 5: , where Fi = {fk, fw, . . . , fW } ⊆ F denotes the set of factor nodes neighboring xi . . . . xi (xi) µxi→fk fk fW fw (xi) µfw→xi (xi) µfW →xi [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Marginal inference of the variable node xi. From (23), the marginal of xi is also a multivariate Gaussian distribution: p(xi) ∝ exp  − 1 2 kxi − xˆik 2 Λxi  , (24) with precision matrix Λxi and mean vector xˆi given by: Λxi = X fa∈Fi Λfa→xi (25a) xˆi = Λ −1 xi X fa∈Fi Λfa→xi zfa→xi . (25b) Finally, the mean xˆi is taken as the estimate of the local state vector xi . Example 5 (Marginal Inference). Margin… view at source ↗
Figure 8
Figure 8. Figure 8: Messages µfk→xi (xi) from factor nodes to variable nodes required for marginal computation. Consider variable node x3. The incoming messages re￾quired to compute its marginal are zf4→x3 and zf6→x3 , with corresponding precision matrices Λf4→x3 and Λf6→x3 . Using (25), the precision matrix and mean vector of the marginal are: Λx3 = Λf4→x3 + Λf6→x3 (26a) xˆ3 = Λ −1 x3 (Λf4→x3 zf4→x3 + Λf6→x3 zf6→x3 ). (26b) … view at source ↗
Figure 9
Figure 9. Figure 9: Messages µfk→xi (xi) from pairwise factor nodes to variable nodes. Consider the fused current phasor measurements associated with factor node f7. The measurement values are stacked into a single vector zf7 = [0.24, −0.67, −0.24, 0.67]T , with the corresponding precision matrix Λf7 = 105 · diag(0.22, 1.46, 0.22, 1.47). The associated coefficient matri￾ces are given by: Hf7,x1 =     2.0 4.0 −4.0 2.0 −2.0… view at source ↗
Figure 10
Figure 10. Figure 10: Normalized RMSE of the scalar, multivariate, and fu [PITH_FULL_IMAGE:figures/full_fig_p009_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: (a). In each simulation, the voltage magnitude and angle variances are randomly selected from σ 2 ∈ [10−8 , 10−6 ], while the current magnitude and angle variances are fixed at σ 2 = 10−6 . This setting examines convergence under different levels of unary factor node precision, including the case where unary and pairwise factor nodes have the same precision. All plotted values of 1 − ρ(Q) remain positive,… view at source ↗
Figure 12
Figure 12. Figure 12: Component-wise absolute error of the fusion-based [PITH_FULL_IMAGE:figures/full_fig_p011_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Voltage magnitude and angle at bus 430 (subfigure a) a [PITH_FULL_IMAGE:figures/full_fig_p011_13.png] view at source ↗
read the original abstract

Electric power systems require accurate, scalable, distributed, and near real-time state estimation (SE) to support reliable monitoring and control under increasingly complex operating conditions. Limited monitoring capabilities can lead to inefficient operation and, in extreme cases, large-scale disturbances such as blackouts. To address these challenges, this paper proposes a vectorized Gaussian belief propagation (GBP) framework for phasor measurement unit-based SE, formulated over factor graphs and specifically designed to support distributed and near real-time monitoring. The proposed framework includes multivariate and fusion-based GBP formulations. The multivariate formulation jointly models related state variables and their measurement relationships, while the fusion-based formulation reduces factor graph complexity by combining multiple measurements associated with the same set of variables, resulting in a structure that more closely reflects the underlying electrical coupling of the power system. The resulting algorithms operate in a fully distributed manner at the bus level and achieve fast convergence and high estimation accuracy, often within a few iterations, as demonstrated by numerical results on systems ranging from 60 to 13659 buses, where the fusion-based formulation achieves single-digit millisecond iteration times on the largest test case.

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

0 major / 3 minor

Summary. The paper proposes a vectorized Gaussian belief propagation (GBP) framework for phasor measurement unit (PMU)-based state estimation (SE) in electric power systems. Formulated over factor graphs, it introduces multivariate GBP (jointly modeling related state variables and measurements) and fusion-based GBP (combining measurements to reflect electrical coupling and reduce complexity). The algorithms are claimed to operate fully distributed at the bus level, achieving fast convergence (often within a few iterations) and high accuracy, with numerical validation on systems from 60 to 13,659 buses and single-digit millisecond iteration times for the fusion-based version on the largest case.

Significance. If the performance claims hold, the work offers a scalable, fully distributed approach to near real-time PMU-based SE without central coordination. This addresses key needs in modern power systems for handling complexity and limited monitoring, potentially enabling reliable grid monitoring at scale. The reported results on very large test cases (up to 13k buses) with low iteration times represent a practical strength for distributed implementations.

minor comments (3)
  1. The abstract and introduction claim 'high estimation accuracy' and 'fast convergence' but do not specify quantitative thresholds (e.g., maximum voltage angle error or iteration count for convergence) used to define these terms; these should be stated explicitly when presenting the numerical results.
  2. Notation for the factor graph construction and message passing (multivariate vs. fusion-based) should include a clear diagram or pseudocode in the methods section to distinguish the two formulations and their computational complexity.
  3. The paper should clarify how the linear PMU measurement model is exactly mapped to the Gaussian factor nodes, including any assumptions on noise covariance and observability.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary of our vectorized Gaussian belief propagation framework for PMU-based state estimation and for recommending minor revision. No specific major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper introduces a new vectorized Gaussian belief propagation algorithm for distributed PMU-based state estimation, formulated as multivariate and fusion-based message passing on factor graphs. The central claims of fully distributed bus-level operation, fast convergence, and accuracy are supported by direct numerical validation across test systems (60 to 13659 buses) rather than by any derivation that reduces to fitted parameters, self-definitional equations, or load-bearing self-citations. The Gaussian factor-graph model is an explicit modeling choice whose outputs are compared against ground-truth simulations, keeping the derivation chain self-contained and externally falsifiable.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the central claim rests on standard domain assumptions about modeling power systems with factor graphs and Gaussian beliefs; no free parameters or invented entities are explicitly introduced in the provided text.

axioms (1)
  • domain assumption Power system state estimation can be accurately modeled using factor graphs with Gaussian probability distributions for states and measurements.
    Invoked implicitly when formulating the multivariate and fusion-based GBP over the electrical network.

pith-pipeline@v0.9.0 · 5519 in / 1288 out tokens · 44958 ms · 2026-05-10T14:53:58.787688+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

31 extracted references · 31 canonical work pages

  1. [1]

    Challenges of variable energy resource in tegration and power system security: Lessons from the 2025 Iberian system blackout,

    M. Gonzalez, “Challenges of variable energy resource in tegration and power system security: Lessons from the 2025 Iberian system blackout,” TESEA, vol. 6, no. 1, pp. 1–6, 2025

    work page 2025

  2. [2]

    A distributed multiarea state estimation ,

    G. N. Korres, “A distributed multiarea state estimation ,” IEEE Trans. Power Syst., vol. 26, no. 1, pp. 73–84, 2010

    work page 2010

  3. [3]

    5G mobile cellular networks: Enabling dist ributed state estimation for smart grids,

    M. Cosovic, A. Tsitsimelis, D. Vukobratovic, J. Matamor os, and C. Anton-Haro, “5G mobile cellular networks: Enabling dist ributed state estimation for smart grids,” IEEE Commun. Mag. , vol. 55, no. 10, pp. 62–69, 2017

    work page 2017

  4. [4]

    Power system real-time monitoring by usin g PMU- based robust state estimation method,

    J. Zhao, G. Zhang, K. Das, G. N. Korres, N. M. Manousakis, A . K. Sinha, and Z. He, “Power system real-time monitoring by usin g PMU- based robust state estimation method,” IEEE Trans. Smart Grid , vol. 7, no. 1, pp. 300–309, 2016

    work page 2016

  5. [5]

    Latest trends in PMU placement techniques,

    H. Sudheendra, V . Kumar Jadoun, N. Jayalakskmi, and A. Ag arwal, “Latest trends in PMU placement techniques,” in Proc. PARC , 2020, pp. 340–345

    work page 2020

  6. [6]

    A secure power syste m distributed state estimation via a consensus-based mechanism and a coop erative trust management strategy,

    S. Nasiri, H. Seifi, and H. Delkhosh, “A secure power syste m distributed state estimation via a consensus-based mechanism and a coop erative trust management strategy,” IEEE Trans Ind. Informat. , vol. 20, no. 2, pp. 3002–3014, 2023

    work page 2023

  7. [7]

    Multi area state estimation using sy nchronized phasor measurements,

    L. Zhao and A. Abur, “Multi area state estimation using sy nchronized phasor measurements,” IEEE Trans. Power Syst. , vol. 20, no. 2, pp. 611–617, 2005

    work page 2005

  8. [8]

    Distributed weighted least-squares estimation for power networks,

    D. Marelli, B. Ninness, and M. Fu, “Distributed weighted least-squares estimation for power networks,” Proc. IF AC-PapersOnLine SYSID , vol. 48, no. 28, pp. 562–567, 2015

    work page 2015

  9. [9]

    Gra dient- based multi-area distribution system state estimation,

    X. Zhou, Z. Liu, Y . Guo, C. Zhao, J. Huang, and L. Chen, “Gra dient- based multi-area distribution system state estimation,” IEEE Trans. Smart Grid , vol. 11, no. 6, pp. 5325–5338, 2020

    work page 2020

  10. [10]

    Real-time distribution system state estimation with asynchronous me asurements,

    G. Cavraro, J. Comden, E. Dall’Anese, and A. Bernstein, “Real-time distribution system state estimation with asynchronous me asurements,” IEEE Trans. Smart Grid , vol. 13, no. 5, pp. 3813–3822, 2022

    work page 2022

  11. [11]

    Synchrophaso r-based real-time state estimation and situational awareness syst em for power system operation,

    H. Chen, L. Zhang, J. Mo, and K. E. Martin, “Synchrophaso r-based real-time state estimation and situational awareness syst em for power system operation,” J. Mod. Power Syst. Clean Energy , vol. 4, no. 3, pp. 370–382, 2016

    work page 2016

  12. [12]

    Dynamic state estimation for power system control and prot ection,

    Y . Liu, A. K. Singh, J. Zhao, A. S. Meliopoulos, B. Pal, M. A. bin Mohd Ariff, T. V an Cutsem, M. Glavic, Z. Huang, I. Kamwa et al. , “Dynamic state estimation for power system control and prot ection,” IEEE Trans. Power Syst. , vol. 36, no. 6, pp. 5909–5921, 2021

    work page 2021

  13. [13]

    Near real -time distributed state estimation via AI/ML-empowered 5G netwo rks,

    O. Kundacina, M. Forcan, M. Cosovic, D. Raca, M. Dzafera gic, D. Miskovic, M. Maksimovic, and D. Vukobratovic, “Near real -time distributed state estimation via AI/ML-empowered 5G netwo rks,” in Proc. SmartGridComm. IEEE, 2022, pp. 284–289

    work page 2022

  14. [14]

    Distributed multi-area state estima tion for power systems with switching communication graphs,

    J. Wang and T. Li, “Distributed multi-area state estima tion for power systems with switching communication graphs,” IEEE Trans. Smart Grid, vol. 12, no. 1, pp. 787–797, 2020

    work page 2020

  15. [15]

    Distributed Gauss–Ne wton method for state estimation using belief propagation,

    M. Cosovic and D. Vukobratovic, “Distributed Gauss–Ne wton method for state estimation using belief propagation,” IEEE Trans. Power Syst. , vol. 34, no. 1, pp. 648–658, 2019

    work page 2019

  16. [16]

    A compl ex domain Gaussian belief propagation method for fully distri buted state estimation,

    K. Sun, Z. Wei, V . Dinavahi, M. Huang, and G. Sun, “A compl ex domain Gaussian belief propagation method for fully distri buted state estimation,” IEEE Trans. Power Syst., vol. 40, no. 1, pp. 982–995, 2024

    work page 2024

  17. [17]

    Distributed weighted leas t-squares estimation for networked systems with edge measurements,

    Q. Y ang, Z. Zhang, and M. Fu, “Distributed weighted leas t-squares estimation for networked systems with edge measurements,” Automatica, vol. 120, p. 109091, 2020

    work page 2020

  18. [18]

    O bservability analysis for large-scale power systems using factor graphs ,

    M. Cosovic, M. Delalic, D. Raca, and D. Vukobratovic, “O bservability analysis for large-scale power systems using factor graphs ,” IEEE Trans. Power Syst., vol. 36, no. 5, pp. 4791–4799, 2021

    work page 2021

  19. [19]

    Power systems obser vability analysis based on parallel Gaussian belief propagation,

    C. Ren, X. Wang, C. Cao, and J. Chen, “Power systems obser vability analysis based on parallel Gaussian belief propagation,” i n Proc. ICC- SNT. IEEE, 2022, pp. 19–23

    work page 2022

  20. [20]

    A Gaussi an belief propagation-based cyber-attack detection approach for st ate estimation model in unbalanced distribution networks,

    S. Wei, Z. Wu, M. Shahidehpour, J. Xu, and Q. Hu, “A Gaussi an belief propagation-based cyber-attack detection approach for st ate estimation model in unbalanced distribution networks,” IEEE Trans. Smart Grid , 2026

    work page 2026

  21. [21]

    Modelling and analysis of electric powe r systems,

    G. Andersson, “Modelling and analysis of electric powe r systems,” ETH Zurich, pp. 5–6, 2008

    work page 2008

  22. [22]

    On the use of PMUs in power system state estimation,

    A. Gomez-Exposito, A. Abur, P . Rousseaux, A. de la Villa Jaen, and C. Gomez-Quiles, “On the use of PMUs in power system state estimation,” in Proc. PSCC, 2011

    work page 2011

  23. [23]

    I. ISO. and B. OIML, Guide to the Expression of Uncertainty in Measurement. Aenor Madrid, Spain, 1993

    work page 1993

  24. [24]

    Abur and A

    A. Abur and A. Exp´ osito, Power System State Estimation: Theory and Implementation, ser. Power Engineering. Taylor & Francis, 2004

    work page 2004

  25. [25]

    Convergen ce analysis of distributed inference with vector-valued Gaussian beli ef propagation,

    J. Du, S. Ma, Y .-C. Wu, S. Kar, and J. M. Moura, “Convergen ce analysis of distributed inference with vector-valued Gaussian beli ef propagation,” Journal of Machine Learning Research, vol. 18, no. 172, pp. 1–38, 2018

    work page 2018

  26. [26]

    Im- pact of inductive current transformers on synchrophasor me asurement in presence of modulations,

    G. D’Avanzo, A. Delle Femine, D. Gallo, C. Landi, and M. L uiso, “Im- pact of inductive current transformers on synchrophasor me asurement in presence of modulations,” Measurement, vol. 155, p. 107535, 2020

    work page 2020

  27. [27]

    J uliagrid: An open-source julia-based framework for power system stat e estima- tion,

    M. Cosovic, O. Kundacina, M. Delalic, A. Teskeredzic, D . Raca, A. Mesanovic, D. Miskovic, D. Vukobratovic, and A. Monti, “J uliagrid: An open-source julia-based framework for power system stat e estima- tion,” SEGAN, p. 102073, 2025

    work page 2025

  28. [28]

    Optimal placement of pmus by integer linear pro gramming,

    B. Gou, “Optimal placement of pmus by integer linear pro gramming,” IEEE Trans. Power Syst. , vol. 23, no. 3, pp. 1525–1526, 2008

    work page 2008

  29. [29]

    Con- tingency ranking with respect to overloads in very large pow er systems taking into account uncertainty, preventive, and corrective actions,

    S. Fliscounakis, P . Panciatici, F. Capitanescu, and L. Wehenkel, “Con- tingency ranking with respect to overloads in very large pow er systems taking into account uncertainty, preventive, and corrective actions,” IEEE Trans. Power Syst. , vol. 28, no. 4, pp. 4909–4917, 2013

    work page 2013

  30. [30]

    Suppressing ineffective control act ions in optimal power flow problems,

    F. Capitanescu, “Suppressing ineffective control act ions in optimal power flow problems,” IET Gener . Transm. Distrib. , vol. 14, no. 13, pp. 2520–2527, 2020

    work page 2020

  31. [31]

    Grid structural characteristics as validation criteria f or synthetic net- works,

    A. B. Birchfield, T. Xu, K. M. Gegner, K. S. Shetye, and T. J . Overbye, “Grid structural characteristics as validation criteria f or synthetic net- works,” IEEE Trans. Power Syst. , vol. 32, no. 4, pp. 3258–3265, 2016

    work page 2016