pith. sign in

arxiv: 2604.05334 · v1 · submitted 2026-04-07 · 📡 eess.SY · cs.SY

CT Saturation Detection and Compensation: A Hybrid Physical Model- and Data-Driven Method

Songhao Yang , Yubo Zhang , Zhiguo Hao , Zexuan Lin , Baohui Zhang This is my paper

Pith reviewed 2026-05-10 19:55 UTC · model grok-4.3

classification 📡 eess.SY cs.SY
keywords current transformer saturationsaturation detectionwaveform compensationhybrid methodfully convolutional networkshort-circuit current modelrelay protectionpower system
0
0 comments X

The pith

A hybrid neural-physical method detects current transformer saturation and reconstructs the true waveform without manual threshold tuning.

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

The paper proposes a hybrid approach combining a Fully Convolutional Network with a physical short-circuit current model to detect and compensate for CT saturation in power systems. The network identifies saturation from waveform data while the physical model reconstructs the undistorted current once saturation is flagged. This integration is intended to remove the need for adjusting universal thresholds and parameters that burden existing techniques. It also aims to improve the generalization and interpretability shortcomings of purely data-driven methods. Simulation and experimental results are presented to support effectiveness and explore uses in relay protection.

Core claim

The method first employs a Fully Convolutional Network trained on simulation and experimental data to classify the operating status of the current transformer as saturated or unsaturated. Upon detection of saturation, a concise physical model of the short-circuit current is applied to compensate the measured waveform and recover the real primary current. The hybrid structure avoids arduous universal threshold adjustments while mitigating generalization and interpretability deficits of data-driven approaches alone.

What carries the argument

The hybrid integration where a Fully Convolutional Network performs saturation detection and a short-circuit current physical model performs waveform compensation.

If this is right

  • Relay protection devices can operate more reliably during faults by receiving compensated undistorted current signals.
  • Saturation handling no longer requires case-by-case threshold or parameter tuning across different CTs and system conditions.
  • The physical model component supplies interpretability that pure neural methods lack, aiding diagnosis of protection events.
  • The approach can be applied to future digital protection schemes that require accurate secondary current reproduction.

Where Pith is reading between the lines

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

  • The same detection-plus-physical-model pattern could apply to other sensor nonlinearities in power systems where a usable physical reconstruction equation exists.
  • If the network generalizes, protection systems might experience fewer false trips or delayed operations during severe faults.
  • The hybrid structure may ease regulatory or safety-case acceptance because the compensation step remains traceable to established circuit equations.

Load-bearing premise

The neural network trained on limited data will correctly detect saturation under all unseen operating conditions, and the physical short-circuit model will accurately reconstruct the true waveform once saturation is flagged.

What would settle it

A real-world test case with operating conditions outside the training distribution where the method either misses saturation or produces a compensated waveform that deviates substantially from independent high-accuracy measurements.

Figures

Figures reproduced from arXiv: 2604.05334 by Baohui Zhang, Songhao Yang, Yubo Zhang, Zexuan Lin, Zhiguo Hao.

Figure 11
Figure 11. Figure 11: CT saturation hardware experiment [PITH_FULL_IMAGE:figures/full_fig_p008_11.png] view at source ↗
Figure 14
Figure 14. Figure 14: The time-delay caused by the proposed method in relay protection. 1  : length of data window,  2 : updating period of data window 3  : time delay of the proposed method, Tdelay: delay time of backup protection [PITH_FULL_IMAGE:figures/full_fig_p009_14.png] view at source ↗
read the original abstract

Current transformer (CT) saturation is one of the dominant causes of relay protection devices' malfunctions, which pose a threat to the safe operation of the power system. To address this problem, we propose a hybrid physical model- and data-driven method. The method firstly detects the CT saturation and then compensates it to reproduce the real waveform. Considering the multi-factor and strong nonlinearity of CT saturation, a data-driven model, namely the Fully Convolutional Network (FCN), is built to detect the operation status of CT. As for the compensation, a physical model of short-circuit current is used for its conciseness and universality. Through tactfully integrating the data model and the physical model, the proposed method is endowed with two major merits: the arduous adjustment of universal thresholds and parameters in existing methods is avoided, and the deficiency in generalization and interpretability of the data-driven method is assuaged. Simulation and experimental results verify the effectiveness of the proposed method. Furthermore, its application potential to future protection is explored.

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 a hybrid physical model- and data-driven method for detecting and compensating current transformer (CT) saturation. A Fully Convolutional Network (FCN) detects saturation from multi-factor nonlinear behavior, after which a concise physical model of short-circuit current reconstructs the true waveform. The integration is claimed to eliminate arduous threshold/parameter tuning required by existing methods while mitigating the generalization and interpretability shortcomings of pure data-driven approaches. Effectiveness is asserted via simulation and experimental results, with potential extension to future protection schemes.

Significance. If the hybrid method demonstrates reliable detection and accurate compensation across operating conditions, it would address a key source of relay malfunctions in power systems. Combining data-driven flexibility for detection with a universal physical reconstruction step could reduce reliance on manual tuning while retaining interpretability, offering a practical advance for protection engineering.

major comments (3)
  1. [Abstract] Abstract: the statement that 'simulation and experimental results verify the effectiveness' provides no quantitative metrics (detection accuracy, compensation RMSE, false-positive rates, or baseline comparisons), error bars, or cross-validation details, leaving the central claims of improved generalization and avoidance of tuning unsupported by evidence.
  2. [FCN detection component] FCN detection section: training occurs on limited simulation plus experimental data without reported out-of-distribution testing across CT ratios, burdens, or inception angles; because compensation is applied only after the FCN flags saturation, any undetected cases directly degrade waveform reconstruction and undermine the claimed generalization merit.
  3. [Compensation and integration] Hybrid integration: the physical short-circuit current model is asserted to be universal and parameter-free once triggered, yet no sensitivity analysis shows how FCN misclassifications (false negatives or positives) propagate to overall error; this conditional dependence makes the 'tactful integration' claim load-bearing but unverified.
minor comments (2)
  1. [Results] Add explicit dataset statistics (number of cases, range of parameters) and a table comparing the proposed method against at least two published baselines on the same test cases.
  2. [Notation and equations] Clarify notation for secondary burden and CT ratio to ensure consistency between the physical model equations and the FCN input features.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We have carefully reviewed each major comment and will revise the paper to address the concerns about quantitative evidence, generalization testing, and sensitivity analysis while preserving the core contributions of the hybrid method.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the statement that 'simulation and experimental results verify the effectiveness' provides no quantitative metrics (detection accuracy, compensation RMSE, false-positive rates, or baseline comparisons), error bars, or cross-validation details, leaving the central claims of improved generalization and avoidance of tuning unsupported by evidence.

    Authors: We agree that the abstract would be strengthened by including quantitative metrics. In the revised version, we will incorporate specific performance figures such as detection accuracy, compensation RMSE, false-positive rates, and baseline comparisons drawn from the simulation and experimental results already presented in the manuscript body. revision: yes

  2. Referee: [FCN detection component] FCN detection section: training occurs on limited simulation plus experimental data without reported out-of-distribution testing across CT ratios, burdens, or inception angles; because compensation is applied only after the FCN flags saturation, any undetected cases directly degrade waveform reconstruction and undermine the claimed generalization merit.

    Authors: The training data was generated to cover a variety of CT ratios, burdens, and inception angles through both simulation and laboratory experiments. We acknowledge that explicit out-of-distribution testing is not separately reported. In the revision, we will add a dedicated evaluation subsection with additional OOD test cases to better substantiate the generalization claims. revision: yes

  3. Referee: [Compensation and integration] Hybrid integration: the physical short-circuit current model is asserted to be universal and parameter-free once triggered, yet no sensitivity analysis shows how FCN misclassifications (false negatives or positives) propagate to overall error; this conditional dependence makes the 'tactful integration' claim load-bearing but unverified.

    Authors: We recognize the value of quantifying the effect of detection errors on compensation performance. The revised manuscript will include a sensitivity analysis that examines error propagation under controlled rates of false negatives and positives, demonstrating the robustness of the hybrid integration. revision: yes

Circularity Check

0 steps flagged

No significant circularity; hybrid detection and compensation remain independent

full rationale

The paper's core chain consists of training an FCN on simulation/experimental data for saturation detection followed by application of an independent short-circuit current physical model for compensation. No equations or steps reduce by construction to fitted inputs, no self-citations are load-bearing, and no uniqueness theorems or ansatzes are smuggled in. The claimed merits (avoiding threshold tuning, improved interpretability) are design assertions verified externally on held-out simulations and experiments rather than tautological re-statements of the training data. This is the normal non-circular outcome for a hybrid ML-plus-physics method.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that a trained FCN can reliably classify saturation states and that a standard short-circuit current equation suffices for waveform reconstruction; no new entities are postulated.

free parameters (1)
  • FCN network weights and hyperparameters
    Learned from training data to perform saturation detection.
axioms (1)
  • domain assumption The short-circuit current physical model accurately represents the unsaturated waveform under fault conditions
    Invoked to justify the compensation step after detection.

pith-pipeline@v0.9.0 · 5489 in / 1227 out tokens · 25155 ms · 2026-05-10T19:55:05.436072+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

38 extracted references · 38 canonical work pages

  1. [1]

    Complete model dev elopment for an optical current transformer,

    S. Kucuksari and G. G. Karady, “Complete model dev elopment for an optical current transformer,” IEEE Trans. Power Del. , vol. 27, no. 4, pp. 1755–1762, Oct. 2012

    work page 2012

  2. [2]

    Evaluation of flexible Rogowski coil performances in power frequency applications,

    M. Chiampi, G. Crotti, and A. Morando, “Evaluation of flexible Rogowski coil performances in power frequency applications,” IEEE Trans. Instrum. Meas., vol. 60, no. 3, pp. 854–862, Mar. 2011

    work page 2011

  3. [3]

    IEEE Committee Report, Gapped core current transformer characteristics and performance,

    B. Bozoki, H. J. Calhoun, and C. M. Gadsden et al., “IEEE Committee Report, Gapped core current transformer characteristics and performance,” IEEE Trans. Power Del., vol. 5, no. 4, pp. 1732–1740, Oct. 1990

    work page 1990

  4. [4]

    IEEE Guide for the Application of Current Transformers Used for Protective Relaying Purposes, IEEE Standard C37.110, 2007

    work page 2007

  5. [5]

    Influence of the number of core air gaps on transient state parameters of TPZ class protective current transformers,

    E. Lesniewska and W. Jalmuzny, “Influence of the number of core air gaps on transient state parameters of TPZ class protective current transformers,” Inst. Eng. Technol. Sci., Meas. Technol., pp. 105–112, Mar. 2009

    work page 2009

  6. [6]

    A saturation suppression approach for the current transformer —Part I: Fundamental concepts and design,

    M. Davarpanah, M. Sanaye -Pasand, and R. Iravani, “A saturation suppression approach for the current transformer —Part I: Fundamental concepts and design,” IEEE Trans. Power De l., vol. 28, no. 3, pp. 1928-1935, Jul. 2013

    work page 1928

  7. [7]

    Current -transformer saturation prevention using a controlled voltage -source compensator,

    E. Hajipour, M. Vakilian, and M. Sanaye -Pasand, “Current -transformer saturation prevention using a controlled voltage -source compensator,” IEEE Trans. Power Del., vol. 32, no. 2, pp. 1039–1048, Apr. 2017

    work page 2017

  8. [8]

    Current Transformer Saturation Compensator by Using Negative Voltage Feedback,

    M. De lzendeh and H. Kazemi Karegar, “Current Transformer Saturation Compensator by Using Negative Voltage Feedback, ” IEEE Trans. on Power Del., vol. 35, no. 3, pp. 1200-1208, June 2020

    work page 2020

  9. [9]

    An algorithm for compensating secondary currents of current transformers,

    Y. C. Kang, J. K. Park, S. H. Kang, A. T. Johns, and R. K. Aggarwal, “An algorithm for compensating secondary currents of current transformers,” IEEE Trans. Power Del., vol. 12, no. 1, pp. 116–124, Jan. 1997

    work page 1997

  10. [10]

    A digital compensation method for improving current transformer accuracy,

    N. Locci and C. Muscas, “A digital compensation method for improving current transformer accuracy,” IEEE Trans. on Power Del., vol. 15, no. 4, pp. 1104-1109, Oct. 2000

    work page 2000

  11. [11]

    Hysteresis and eddy currents compensation in current transformers,

    N. Locci and C. Muscas, “Hysteresis and eddy currents compensation in current transformers,” IEEE Trans. on Power Del. , vol. 16, no. 2, pp. 154-159, April 2001

    work page 2001

  12. [12]

    Compensation of the distortion in the secondary current caused by saturation and remanence in a CT,

    Yong Cheol Kang, Ui Jai Lim, Sang Hee Kang and P. A. Crossley, “Compensation of the distortion in the secondary current caused by saturation and remanence in a CT,” IEEE Trans. on Power Del., vol. 19, no. 4, pp. 1642-1649, Oct. 2004

    work page 2004

  13. [13]

    Current-Transformer Saturation Compensation for Transformer Differential Relays,

    E. Hajipour, M. Vakilian and M. Sanaye -Pasand, “Current-Transformer Saturation Compensation for Transformer Differential Relays, ” IEEE Trans. on Power Del., vol. 30, no. 5, pp. 2293-2302, Oct. 2015

    work page 2015

  14. [14]

    Correction of saturated current from measurement current transformer,

    Ozgonenel, O., “Correction of saturated current from measurement current transformer,” IET Electr. Power Appl. , vol. 7, no. 7, pp. 580-585, Aug. 2013

    work page 2013

  15. [15]

    An efficient compensation algorithm for current transformer saturation effects,

    Jiuping Pan, Khoi Vu and Yi Hu, “An efficient compensation algorithm for current transformer saturation effects, ” IEEE Trans. on Power Del. , vol. 19, no. 4, pp. 1623-1628, Oct. 2004

    work page 2004

  16. [16]

    Development of a new derivative-based algorithm to detect current transformer saturation,

    A. Hooshyar, M. Sanaye-Pasand, and M. Davarpanah, “Development of a new derivative-based algorithm to detect current transformer saturation,” IET Gen., Transm. Distrib., vol. 6, no. 3, pp. 207–217, 2012

    work page 2012

  17. [17]

    A new method of current -transformer saturation detection in the presence of noise,

    B. Schettino, C. Duque, P. Silveira, P. Ribeiro, and A. Cerqueira, “A new method of current -transformer saturation detection in the presence of noise,” IEEE Trans. Power Del. , vol. 29, no. 4, pp. 1760 –1767, Aug. 2014

    work page 2014

  18. [18]

    A CT saturation detection algorithm,

    Y. C. Kang, S. H. Ok, and S. H. Kang, “A CT saturation detection algorithm,” IEEE Trans. on Power Del., vol. 19, no. 1, pp. 78 –85, Jan. 2004

    work page 2004

  19. [19]

    Compensation of distorted secondary current caused by saturation and remanence in a current transformer,

    Y. Y. Hong and D. W. Wei, “Compensation of distorted secondary current caused by saturation and remanence in a current transformer,” IEEE Trans. Power Del., vol. 25, no. 1, pp. 47–54, Jan. 2010

    work page 2010

  20. [20]

    A wavelet-based transformer differential protection with differential current transformer saturation and crosscountry fault detection,

    R. P. Medeiros and F. B. Costa, “A wavelet-based transformer differential protection with differential current transformer saturation and crosscountry fault detection,” IEEE Trans. Power Del., vol. 33, no. 2, pp. 789–799, Apr. 2018

    work page 2018

  21. [21]

    Fast Detection and Compensation of Current Transformer Saturation Using Extended Kalman Filter,

    F. Naseri, Z. Kazemi, E. Farjah and T. Ghanbari, “Fast Detection and Compensation of Current Transformer Saturation Using Extended Kalman Filter,” IEEE Trans. on Power Del., vol. 34, no. 3, pp. 1087-1097, June 2019

    work page 2019

  22. [22]

    A series multiresolution morphological gradient-based criteri on to identify CT saturation,

    Xiangning Lin, Li Zou, Qing Tian, Hanli Weng and Pei Liu, “A series multiresolution morphological gradient-based criteri on to identify CT saturation,” IEEE Trans. on Power Del. , vol. 21, no. 3, pp. 1169 -1175, July 2006

    work page 2006

  23. [23]

    Current transformer saturation detection using morphological gradient and morphological decomposition and its hardware implementation,

    T. Ji, M. Shi, M. Li, L. Zhang, and Q. Wu, “Current transformer saturation detection using morphological gradient and morphological decomposition and its hardware implementation,” IEEE Trans. Ind. Electron., vol. 64, no. 6, pp. 4721–4729, Jun. 2017

    work page 2017

  24. [24]

    Correction of current transformer distorted secondary currents due to saturation using artificial neural networks,

    D. C. Yu and J. C. Cummins, “Correction of current transformer distorted secondary currents due to saturation using artificial neural networks,” IEEE Trans. Power Del., vol. 16, no. 2, pp. 189–194, Apr. 2001

    work page 2001

  25. [25]

    Compensation of CT distorted secondary current waveform in online conditions,

    F. Haghjoo and M. Hadi Pak, “Compensation of CT distorted secondary current waveform in online conditions,” IEEE Trans. Power Del., vol. 31, no. 2, pp. 711–720, Apr. 2016

    work page 2016

  26. [26]

    Integrating model -driven and datadriven methods for power system frequency stability assessment and control,

    Q. Wang, F. Li, Y. Tang, and Y. Xu, “Integrating model -driven and datadriven methods for power system frequency stability assessment and control,” IEEE Trans. Power Syst. , vol. 34, no. 6, pp. 4557 –4568, Nov. 2019

    work page 2019

  27. [27]

    A New Adaptive Dependability-Security Approach to Enhance Wide Area Back -Up Protection of Transmission System,

    M. K. Jena, S. R. Samantaray and B. K. Panigrahi, “A New Adaptive Dependability-Security Approach to Enhance Wide Area Back -Up Protection of Transmission System,” IEEE Trans. on Smart Grid , vol. 9, no. 6, pp. 6378-6386, Nov. 2018

    work page 2018

  28. [28]

    Exploring Key Weather Factors From Analytical M odeling Toward Improved Solar Power Forecasting,

    J. Wang, H. Zhong, X. Lai, Q. Xia, Y. Wang and C. Kang, “Exploring Key Weather Factors From Analytical M odeling Toward Improved Solar Power Forecasting, ” IEEE Trans. on Smart Grid, vol. 10, no. 2, pp. 1417-1427, March 2019

    work page 2019

  29. [29]

    Combining Data -driven and model-driven methods for robust facial landmark detection,

    H. Zhang, Q. Li, Z. Sun, and Y. Liu, “Combining Data -driven and model-driven methods for robust facial landmark detection,” IEEE Trans. Inf. Forensics Secur., vol. 13, no. 10, pp. 2409–2422, Oct. 2018

    work page 2018

  30. [30]

    Integrating model -driven and data-driven techniques for analyzing learning behaviors in openended learning environments,

    J. S. Kinnebrew, J. R. Segedy, and G. Biswas, “Integrating model -driven and data-driven techniques for analyzing learning behaviors in openended learning environments,” IEEE Trans. Learn. Technol., vol. 10, no. 2, pp. 140–153, Apr.–Jun. 2017. IEEE Transactions on Power Delivery 11

    work page 2017

  31. [31]

    Fully convolutional networks for semantic segmentation,

    E. Shelhamer, J. Long, and T. Darrell, “Fully convolutional networks for semantic segmentation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 39, no. 4, pp. 640–651, Apr. 2017

    work page 2017

  32. [32]

    The Levenber g-Marquardt algorithm: Implementation and theory,

    J. J. Moré, “The Levenber g-Marquardt algorithm: Implementation and theory,” in Numerical Analysis. New York, NY, USA: Springer, 1978, pp. 105–116

    work page 1978

  33. [33]

    Training - and test -time data augmentation for hyperspectral image segmentation,

    J. Nalepa, M. Myller, and M. Kawulok, “Training - and test -time data augmentation for hyperspectral image segmentation,” IEEE Geosci. Remote Sens. Lett., vol. 17, no. 2, pp. 292–296, Feb. 2020

    work page 2020

  34. [34]

    Going deeper with convolutions,

    C. Szegedy et al., “Going deeper with convolutions,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Jun. 2015, pp. 1–9

    work page 2015

  35. [35]

    Theory of ferromagnetic hysteresis,

    D. C. Jiles and D. L. Atherton, "Theory of ferromagnetic hysteresis," J. Magn. Magn. Mater., vol. 61, nos. 1–2, pp. 48–60, 1986

    work page 1986

  36. [36]

    Improved simulation models for current and voltage transformers in relay studies,

    J. R. Lucas, P. G. McLaren, W. W. L. Keerthipala, and R. P. Jayasinghe, “Improved simulation models for current and voltage transformers in relay studies,” IEEE Trans. Power Del., vol. 7, no. 1, pp. 152–159, Jan. 1992

    work page 1992

  37. [37]

    Adaptive Current Differential Protection Schemes for Transmission-Line Protection,

    S. Dambhare, S. A. Soman and M. C. Chandorkar, "Adaptive Current Differential Protection Schemes for Transmission-Line Protection," IEEE Trans. on Power Del., vol. 24, no. 4, pp. 1832-1841, Oct. 2009

    work page 2009

  38. [38]

    In -datacenter performance analysis of a tensor processing unit,

    Jouppi et al., “In -datacenter performance analysis of a tensor processing unit,” in ACM/IEEE 44th Ann. Int. Symp. Comput. Architecture (ISCA) , Toronto, ON, Canada, 2017, pp. 1–12. Songhao Yang. (S’18-M’19) was born in Shandong, China, in 1989. He received the B.S. and Ph.D. degrees in electrical engineering from the Xi’an Jiaotong University, Xi’an, Chi...

    work page 2017