Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 41 of the 41 references it cites were found in a registry or an index; 2 of the 9 sources its claims cite were checked for support.

A Physics-Guided Transformer for Robust State of Charge Estimation in Aging Lithium-Ion Batteries

Open access
Li, Xiang, Wu, Guanru, Chang, Fei, Xia, Weidong, Sun, Shaobin, Shen, Yingjun · 2025 · Batteries

Abstract

Accurate state of charge (SOC) estimation is a critical challenge for battery management systems (BMSs), hindered by the nonlinear electrochemistry of lithium-ion batteries, their sensitivity to temperature, and pervasive measurement noise. Crucially, battery aging significantly degrades estimation accuracy, posing a major hurdle for long-term system dependability. We propose the Physics-Informed Transformer (PI-Transformer), a novel framework that integrates high-fidelity electrochemical constraints from the PyBaMM (Version: 25.10.2) model directly into a Transformer architecture. This approach ensures physical consistency while leveraging the Transformer’s self-attention mechanism to model long-term temporal dependencies. The framework is specifically designed to be robust against the effects of battery aging, incorporating an attention-based noise modeling module to enhance resilience against sensor uncertainty and capacity fade. Evaluated on two public datasets under diverse conditions, including variable temperatures, fast-charging protocols, and multiple stages of battery degradation, the PI-Transformer consistently achieves state-of-the-art performance. It demonstrates exceptional robustness and maintains high accuracy under challenging low-temperature and severely aged battery scenarios, highlighting its strong potential for deployment in real-world ESS applications where aging is a primary concern.

Overview

CoChat Summary

TL;DR

The paper proposes PI-Transformer, a Transformer with PyBaMM-derived physics-informed embedding and attention-based noise/aging modeling for SOC estimation in aging lithium-ion batteries. It reports state-of-the-art RMSE and R² across NASA 24 °C/4 °C and Braatz 30 °C fast-charging/aging datasets.

Key findings

  • On NASA 24 °C (EXP1), PI-Transformer achieved Validation RMSE 0.0195 and Test RMSE 0.0256, with Validation R² 0.9903 and Test R² 0.9765; the paper states it outperformed the next-best LSTM by 3.5% and 25.2% on validation and test RMSE, respectively.
  • On NASA 4 °C (EXP2), it achieved Validation/Test RMSE 0.0226 and R² 0.9783, versus standard Transformer R² 0.9759.
  • On Braatz 30 °C fast-charging/aging (EXP3), it achieved Test RMSE 0.0698, a 31.5% reduction versus CNN-Transformer (0.1016), and Test R² 0.9594.
  • Ablation: disabling the noise modeling module increased RMSE by 13.3% on NASA 4 °C and 18.4% on Braatz; the Conclusion separately states RMSE increased by over 180% when disabled.

Methods

  • PI-Transformer embeds the PyBaMM electrochemical SOC differential equation as a physics-informed embedding, fuses it with data-driven Transformer features, and uses an attention-based noise/aging module; it is trained end-to-end with MSE loss.
  • Data: NASA dataset (34 18650-type cells, 2.0 Ah, 24 °C and 4 °C, 7,457,030 samples across 7565 cycles) and Braatz dataset (124 LFP/graphite A123 APR18650M1A cells, 1.1 Ah, 3.3 V, fast-charging at 30 °C). NASA used a chronological 60/20/20 split per cell; Braatz used a stratified random 60/20/20 split by battery ID.
  • Compared EKF, LSTM, GRU, Transformer, CNN-Transformer, PI-Transformer, and PI-Transformer without noise; metrics were RMSE, R², and MAE. Training used AdamW, learning rate 0.001 with cosine annealing, batch size 1536, 10 epochs with early stopping, PyTorch 2.0/CUDA 12.1 on an NVIDIA A100; inputs were voltage, current, and temperature, with window length 50, d_model 128, 2 encoder layers, 8 heads, d_ff 512, and dropout 0.2.

Limitations

  • Stated: evaluation is limited to Li-ion/LFP chemistries; reliance on accurate temperature measurements may limit practical deployment; computational cost may require optimization for real-time edge inference; and lack of uncertainty quantification limits safety-critical applications.
  • Evident: the ablation magnitude is inconsistent—Section 5.2 reports 13.3% and 18.4% RMSE increases when the noise module is disabled, while the Conclusion says “over 180%.”
  • Evident: Table 3 reports identical validation and test RMSE/R² for PI-Transformer in EXP2 (0.0226/0.9783), which may indicate a reporting or duplication issue.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/batteries11120446

Cites

41 of 41 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

1 Valid · 1 Not valid · 6 Not checked · 66 Own results and interpretations

31 more claims

Create a free account to see all 74 claims and to verify the 1 reference still open.

See all claims — free account
  • It may state two separate claims as one. 29
  • Probably not a substantial claim. 10
  • The quote only partly states this claim. 1

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

5 of 5

Citations

Cites

Show all 35 references

Cited by

See all 2

Metadata

Authors
Li, X., Wu, G., Chang, F., Xia, W., Sun, S., & Shen, Y.
Affiliations
  • State Grid Nanjing Power Supply Company (China) (CN)
  • Chinese Academy of Sciences (CN)
Corresponding author
Wu, Guanru
Venue
Batteries
Publisher
MDPI AG
Published
2025-12-03
Volume
11
Issue
12
Pages
446
ISSN
2313-0105 (ISSN-L)
OpenAlex
W4416958527
Cited by
2
References
35
Citations by year
2026: 2
FWCI
0.49
Type
journal article
Language
en
Open access
Locations
  • Batteries (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Science and Technology Project of State Grid Corporation of China (JF2025021)
  • Incubation Project of State Grid Jiangsu Electric Power Co., Ltd.
Topics
  • Physical Sciences › Engineering › Automotive Engineering › Battery Modeling and Management
  • Physical Sciences › Engineering › Electrical and Electronic Engineering › Advancements in Battery Materials
  • Physical Sciences › Engineering › Control and Systems Engineering › Control Systems and Identification
Keywords
state of charge estimation, lithium-ion batteries, battery degradation, physics-informed machine learning, transformer models, PyBaMM, battery management systems, energy storage systems, capacity fade, temperature effects, fast charging

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-09, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-09 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: A Physics-Guided Transformer for Robust State of Charge Estimation in Aging Lithium-Ion Batteries. https://app.cochat.ai/papers/10.3390/batteries11120446
The paper
Li, X., Wu, G., Chang, F., Xia, W., Sun, S., & Shen, Y. (2025). A Physics-Guided Transformer for Robust State of Charge Estimation in Aging Lithium-Ion Batteries. Batteries. https://doi.org/10.3390/batteries11120446

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.