pith. sign in

arxiv: 2606.27781 · v1 · pith:2RJ274VNnew · submitted 2026-06-26 · 📡 eess.SY · cs.SY

Repair-before-veto control for safe lithium-ion fast charging under unknown ambient and cooling-fault conditions

Yifan Wang This is my paper

Pith reviewed 2026-06-29 03:36 UTC · model grok-4.3

classification 📡 eess.SY cs.SY
keywords fast charginglithium-ion batterysafety controlcooling faultmargin-aware controlbattery managementthermal safetyfault monitoring
0
0 comments X

The pith

A margin-aware controller completes fast charging safely in all nine unknown ambient and cooling conditions while charging 37.9 percent faster than fixed safe currents.

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

The paper establishes that a single controller can manage lithium-ion fast charging when ambient temperature and cooling health are unknown by requesting an aggressive current and then repairing it online to the smallest measured safety margin. Fixed currents either overheat the cell when cooling is degraded or charge too slowly when conditions are favorable, and rigid shutdowns fail to finish the charge. The controller observes terminal voltage, cell temperature, and negative-electrode plating overpotential in real time and reduces current only as needed rather than committing to a schedule or halting. This matters because chargers in the field cannot know each cell's exact thermal state or cooling performance yet must still deliver usable charge without damage. Evaluation in a high-fidelity model across nine combinations of temperature and cooling health shows the approach finishes every case without overheating while producing less plated lithium than alternatives.

Core claim

RACL-B requests an aggressive current and repairs it online to the tightest measured margin among terminal voltage, cell temperature, and negative-electrode lithium-plating overpotential, rather than committing to a fixed schedule or shutting charging down. Under a strict 45.0 °C peak-temperature audit the method safely completes all nine conditions spanning 10/25/40 °C ambient temperature and 100/60/40 % cooling health, is 37.9 % faster than the fastest fixed current safe across the whole envelope, produces the least plated lithium, and remains safe across thermal guard bands. The same margin-aware principle drives a transient-credit fault readout that gives the strongest learned sequence-t

What carries the argument

The repair-before-veto controller (RACL-B) that continuously sets current to the minimum of the three real-time measured margins instead of a preset profile or immediate shutdown.

If this is right

  • The controller finishes charging in every one of the nine tested ambient-temperature and cooling-health combinations without exceeding the 45 °C limit.
  • Charging completes 37.9 % faster than the fastest fixed current that remains safe over the entire set of conditions.
  • The method produces less lithium plating than fixed-current or scheduled protocols while still completing the charge.
  • Safety holds when thermal guard bands are applied around the measured margins.
  • The identical margin principle yields a monitor that localizes cooling-fault onset more accurately than other sequence-to-global methods on real pack data.

Where Pith is reading between the lines

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

  • The separation of an aggressive request from margin-based repair could be applied to other battery constraints if corresponding real-time measurements become available.
  • Real-pack deployment would require confirming that voltage, temperature, and plating-overpotential estimates remain reliable when sensor noise and pack-level effects are present.
  • The same repair logic might reduce risk in grid-storage or second-life battery systems that also face variable cooling performance.
  • Because the controller only reduces current when a margin is approached, it could be combined with separate speed-optimization layers without changing the safety layer.

Load-bearing premise

The high-fidelity model with partially reversible lithium plating and lumped thermal coupling accurately captures how the three measurable margins evolve under unknown cooling degradation so that online repairs keep the cell inside safe limits.

What would settle it

A laboratory test that applies the controller to a physical cell with an introduced cooling fault at 40 °C ambient and records whether peak temperature stays below 45 °C and total plated lithium stays below the model's prediction.

Figures

Figures reproduced from arXiv: 2606.27781 by Yifan Wang.

Figure 1
Figure 1. Figure 1: Overview of the EVENT-BMS framework. The monitoring path and the [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Safe fast charging under unknown ambient [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Guard-band robustness of RACL-B. Strict 9/9 safe completion is preserved across thermal guard bands from 44.5 to 44.85 ◦C (left); only the zero-margin 45.0 ◦C guard overshoots the hard limit (centre); and the average charge time varies only between 62.9 and 64.6 min, well below the safe fixed and reactive baselines (right). 6.4. Transient-aware fault monitoring The monitoring side validates the same margin… view at source ↗
Figure 4
Figure 4. Figure 4: Cross-temperature feasibility under the strict audit. Fixed CC-CV protocols [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Temporal credit dilution in a real faulted window. The load-driven temperature [PITH_FULL_IMAGE:figures/full_fig_p017_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Out-of-distribution cooling-fault detection across three operating-condition shifts [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Credit-dilution theory. The pooled event credit vanishes as [PITH_FULL_IMAGE:figures/full_fig_p019_7.png] view at source ↗
read the original abstract

Fast charging is decisive for electric-vehicle adoption, but field chargers are deployed as one setting while the cell's true thermal state, ambient temperature, and cooling-system health are uncertain. A current that is safe for a healthy cell at room temperature can overheat the same cell when it is hot or its cooling is degraded. We formulate this as a single-setting, unknown-state safe-fast-charging problem and solve it with a margin-aware repair-before-veto controller (RACL-B). RACL-B requests an aggressive current and repairs it online to the tightest measured margin among terminal voltage, cell temperature, and negative-electrode lithium-plating overpotential, rather than committing to a fixed schedule or shutting charging down. We evaluate one deployed setting across nine conditions, spanning 10/25/40 $^\circ$C ambient temperature and 100/60/40\% cooling health, in a high-fidelity Doyle--Fuller--Newman model with partially reversible lithium plating and lumped thermal coupling. Under a strict 45.0 $^\circ$C peak-temperature audit, fixed and ambient-scheduled protocols overheat in five of nine conditions because neither observes hidden cooling degradation, and rigid protective shutdown fails to deliver the charge in every condition. RACL-B safely completes all nine conditions, is 37.9\% faster than the fastest fixed current safe across the whole envelope, produces the least plated lithium, and remains safe across thermal guard bands. The same margin-aware principle drives a transient-credit fault readout (CREST-B) that, on a real introduced-fault battery-pack dataset, gives the strongest learned sequence-to-global monitor for localizing cooling-fault onset under operating-condition shift. The framework provides a deployable thermal-safety guarantee for fast charging together with a margin-aware monitor for the same physical fault class.

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

2 major / 1 minor

Summary. The manuscript presents a repair-before-veto control strategy (RACL-B) for safe lithium-ion battery fast charging under uncertain ambient temperature and cooling system health. The controller dynamically adjusts the charging current based on the minimum margin among terminal voltage, temperature, and lithium plating overpotential to avoid violations while maximizing speed. Simulations in a Doyle-Fuller-Newman model across nine combinations of ambient temperatures (10/25/40°C) and cooling health levels (100/60/40%) demonstrate that RACL-B completes charging safely in all cases, achieves 37.9% faster charging than the safest fixed-current protocol, minimizes lithium plating, and outperforms fixed and ambient-scheduled methods which fail in multiple conditions. Additionally, a related margin-aware monitor (CREST-B) is shown to perform well on a real battery-pack dataset for detecting cooling faults.

Significance. If the observability of the plating overpotential margin holds under real conditions, the work offers a practical single-setting control approach for robust fast charging that provides explicit thermal safety guarantees despite unknown cooling degradation, outperforming static protocols in simulation. The multi-condition evaluation in a high-fidelity DFN model with reversible plating and thermal coupling, plus the CREST-B extension validated on real fault data, strengthens the case for margin-aware methods in EV battery management. The absence of real-cell controller validation and sensitivity analysis tempers immediate impact but does not negate the simulation-based insight.

major comments (2)
  1. [Abstract and §3] Abstract and §3: The RACL-B controller is defined to repair current to the 'tightest measured margin' among terminal voltage, cell temperature, and negative-electrode lithium-plating overpotential. The manuscript provides no sensor model, state estimator, or noise analysis showing how the internal plating overpotential can be obtained from terminal measurements alone when cooling health is unknown and the thermal model is degraded. This is load-bearing for the central claim, as the reported safety in all nine conditions and 37.9% speed-up rest on accurate real-time access to these margins.
  2. [Evaluation across nine conditions (abstract)] Evaluation across nine conditions (abstract): The performance numbers (37.9% faster, safe completion, least plated lithium) are obtained from DFN simulation without reported error bars, sensitivity analysis on margin thresholds, or checks against model mismatch induced by the unknown cooling fault. This weakens the guarantee that the margin-based decisions translate when the plant deviates from the model used to compute the 'measured' margins.
minor comments (1)
  1. The abstract is lengthy and packs multiple claims; consider splitting or condensing the description of RACL-B and CREST-B for improved readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive report. Below we respond point-by-point to the two major comments. We agree that additional discussion is warranted on both observability assumptions and evaluation robustness, and we will incorporate clarifications and new analysis in the revision.

read point-by-point responses

  1. Referee: [Abstract and §3] Abstract and §3: The RACL-B controller is defined to repair current to the 'tightest measured margin' among terminal voltage, cell temperature, and negative-electrode lithium-plating overpotential. The manuscript provides no sensor model, state estimator, or noise analysis showing how the internal plating overpotential can be obtained from terminal measurements alone when cooling health is unknown and the thermal model is degraded. This is load-bearing for the central claim, as the reported safety in all nine conditions and 37.9% speed-up rest on accurate real-time access to these margins.

    Authors: We agree that the manuscript does not supply a sensor model, state estimator, or noise analysis for recovering the lithium-plating overpotential margin from terminal measurements under unknown cooling health. The reported results are obtained inside a perfect-information DFN simulation in which all internal states are directly available; the controller is therefore evaluated under the assumption that the three margins can be measured or estimated in real time. The central contribution is the repair-before-veto logic itself rather than an observer design. We will add an explicit paragraph in §3 stating this modeling assumption and noting that practical realization requires a separate observer (e.g., an extended Kalman filter or neural observer) whose design is left for future work. No change to the numerical results is required. revision: partial

  2. Referee: [Evaluation across nine conditions (abstract)] Evaluation across nine conditions (abstract): The performance numbers (37.9% faster, safe completion, least plated lithium) are obtained from DFN simulation without reported error bars, sensitivity analysis on margin thresholds, or checks against model mismatch induced by the unknown cooling fault. This weakens the guarantee that the margin-based decisions translate when the plant deviates from the model used to compute the 'measured' margins.

    Authors: The nine-condition study is deterministic: the same DFN model supplies both the plant dynamics and the “measured” margins, so model mismatch is deliberately absent by construction. The experiment therefore isolates the effect of unknown ambient temperature and cooling health on a fixed controller. Because the simulations contain no stochastic disturbances, error bars were not reported. We accept that a sensitivity study on the margin thresholds and a brief discussion of potential model mismatch would strengthen the claims. In the revision we will add (i) a sensitivity sweep over the three margin thresholds and (ii) a short paragraph in the discussion section addressing how observer error or thermal-model mismatch could affect closed-loop behavior. revision: yes

Circularity Check

0 steps flagged

No circularity; RACL-B defined from external physical margins evaluated in independent DFN simulation.

full rationale

The paper defines the controller directly from observable margins (voltage, temperature, plating overpotential) and evaluates performance against a high-fidelity DFN plant across nine conditions. No equations, parameters, or predictions reduce the claimed 37.9% speed-up or safety results to quantities fitted or defined by the authors' own prior work. The derivation chain remains self-contained against the stated model benchmarks with no self-citation load-bearing steps or fitted-input predictions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract provides no explicit free parameters, axioms, or invented entities; the controller relies on the standard DFN model assumptions and the existence of online measurements of the three safety margins.

pith-pipeline@v0.9.1-grok · 5860 in / 1327 out tokens · 34113 ms · 2026-06-29T03:36:38.588103+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

18 extracted references · 15 canonical work pages · 1 internal anchor

  1. [1]

    Tomaszewska, Z

    A. Tomaszewska, Z. Chu, X. Feng, S. O’Kane, X. Liu, J. Chen, C. Ji, E. Endler, R. Li, L. Liu, Y. Li, S. Zheng, S. Vetterlein, M. Gao, J. Du, 17 meanlastmax window attn attn_slotsgroup_rwCREST-Boracle 0.4 0.5 0.6 0.7 0.8 0.9 1.0OOD AUROC (mean ± sd) S1: unseen blockage + US06 meanlastmax window attn attn_slotsgroup_rwCREST-Boracle 0.4 0.5 0.6 0.7 0.8 0.9 1...

    work page doi:10.1016/j.etran.2019.100011 2019

  2. [2]

    Waldmann, B.-I

    T. Waldmann, B.-I. Hogg, M. Wohlfahrt-Mehrens, Li plating as un- wanted side reaction in commercial li-ion cells – a review, Journal of Power Sources 384 (2018) 107–124.doi:10.1016/j.jpowsour.2018. 02.063

    work page doi:10.1016/j.jpowsour.2018 2018

  3. [3]

    S. E. J. O’Kane, W. Ai, G. Madabattula, D. Alonso-Alvarez, R. Timms, V. Sulzer, J. S. Edge, B. Wu, G. J. Offer, M. Marinescu, Lithium-ion battery degradation: how to model it, Physical Chemistry Chemical Physics 24 (13) (2022) 7909–7922.doi:10.1039/D2CP00417H

    work page doi:10.1039/d2cp00417h 2022

  4. [4]

    Waldmann, M

    T. Waldmann, M. Wilka, M. Kasper, M. Fleischhammer, M. Wohlfahrt- Mehrens, Temperaturedependentageingmechanismsinlithium-ionbat- teries – a post-mortem study, Journal of Power Sources 262 (2014) 129– 135.doi:10.1016/j.jpowsour.2014.03.112

    work page doi:10.1016/j.jpowsour.2014.03.112 2014

  5. [5]

    D. P. Finegan, J. Zhu, X. Feng, M. Keyser, M. Ulmefors, W. Li, M. Z. Bazant, S. J. Cooper, The application of data-driven methods and physics-based learning for improving battery safety, Joule 5 (2) (2021) 316–329.doi:10.1016/j.joule.2020.12.009. 18 10 1 event fraction 10 3 10 2 10 1 E (a) pooled event credit vanishes as ( 2) E (event credit share) 2 refe...

    work page doi:10.1016/j.joule.2020.12.009 2021

  6. [6]

    K. A. Severson, P. M. Attia, N. Jin, N. Perkins, B. Jiang, Z. Yang, M. H. Chen, M. Aykol, P. K. Herring, D. Fraggedakis, M. Z. Bazant, S. J. Harris, W. C. Chueh, R. D. Braatz, Data-driven prediction of battery cycle life before capacity degradation, Nature Energy 4 (5) (2019) 383– 391.doi:10.1038/s41560-019-0356-8

    work page doi:10.1038/s41560-019-0356-8 2019

  7. [7]

    P. M. Attia, A. Grover, N. Jin, K. A. Severson, T. M. Markov, Y.- H. Liao, M. H. Chen, B. Cheong, N. Perkins, Z. Yang, P. K. Her- ring, M. Aykol, S. J. Harris, R. D. Braatz, S. Ermon, W. C. Chueh, Closed-loop optimization of fast-charging protocols for batteries with machine learning, Nature 578 (7795) (2020) 397–402.doi:10.1038/ s41586-020-1994-5

    2020

  8. [8]

    H. E. Perez, X. Hu, S. Dey, S. J. Moura, Optimal charging of li-ion bat- teries with coupled electro-thermal-aging dynamics, IEEE Transactions on Vehicular Technology 66 (9) (2017) 7761–7770.doi:10.1109/TVT. 2017.2676044

    work page doi:10.1109/tvt 2017

  9. [9]

    Naguib, J

    M. Naguib, J. Chen, P. Kollmeyer, A. Emadi, Thermal fault detection of lithium-ion battery packs through an integrated physics and deep neural network based model, Communications Engineering 4 (2025) 79. doi:10.1038/s44172-025-00409-2

    work page doi:10.1038/s44172-025-00409-2 2025

  10. [10]

    M.Doyle, T.F.Fuller, J.Newman, Modelingofgalvanostaticchargeand 19 discharge of the lithium/polymer/insertion cell, Journal of The Electro- chemical Society 140 (6) (1993) 1526–1533.doi:10.1149/1.2221597

    work page doi:10.1149/1.2221597 1993

  11. [11]

    T. F. Fuller, M. Doyle, J. Newman, Simulation and optimization of the dual lithium ion insertion cell, Journal of The Electrochemical Society 141 (1) (1994) 1–10.doi:10.1149/1.2054684

    work page doi:10.1149/1.2054684 1994

  12. [12]

    Sulzer, S

    V. Sulzer, S. G. Marquis, R. Timms, M. Robinson, S. J. Chapman, Python battery mathematical modelling (PyBaMM), Journal of Open Research Software 9 (1) (2021) 14.doi:10.5334/jors.309

    work page doi:10.5334/jors.309 2021

  13. [13]

    L. Lu, X. Han, J. Li, J. Hua, M. Ouyang, A review on the key issues for lithium-ion battery management in electric vehicles, Journal of Power Sources 226 (2013) 272–288.doi:10.1016/j.jpowsour.2012.10.060

    work page doi:10.1016/j.jpowsour.2012.10.060 2013

  14. [14]

    M.Naguib, P.Kollmeyer, A.Emadi, Applicationofdeepneuralnetworks for lithium-ion battery surface temperature estimation under driving and fast charge conditions, IEEE Transactions on Transportation Elec- trification 9 (1) (2023) 1153–1165.doi:10.1109/TTE.2022.3200225

    work page doi:10.1109/tte.2022.3200225 2023

  15. [15]

    Geirhos, J.-H

    R. Geirhos, J.-H. Jacobsen, C. Michaelis, R. Zemel, W. Brendel, M. Bethge, F. A. Wichmann, Shortcut learning in deep neural net- works, Nature Machine Intelligence 2 (11) (2020) 665–673.doi:10. 1038/s42256-020-00257-z

    2020

  16. [16]

    Invariant Risk Minimization

    M. Arjovsky, L. Bottou, I. Gulrajani, D. Lopez-Paz, Invariant risk min- imization, arXiv preprint arXiv:1907.02893 (2019)

    work page internal anchor Pith review Pith/arXiv arXiv 1907

  17. [17]

    Sagawa, P

    S. Sagawa, P. W. Koh, T. B. Hashimoto, P. Liang, Distributionally robust neural networks for group shifts: On the importance of regu- larization for worst-case generalization, in: International Conference on Learning Representations (ICLR), 2020

    2020

  18. [18]

    Naguib, P

    M. Naguib, P. Kollmeyer, J. Chen, A. Emadi, Battery pack with intro- duced faults dataset – air cooled SBLimotive 5Ah, Borealis Data (2025). doi:10.5683/SP3/THZTJC. 20

    work page doi:10.5683/sp3/thztjc 2025