CoChat

State-of-charge estimation of commercial electric vehicles by using battery-aware transformer network algorithms

Open access
Mu, Bao-Qi, Doukhi, Oualid, Wang, Linfeng, Lee, Deok-Jin · 2025 · Journal of Computational Design and Engineering

The DOI matches its Crossref record; 35 of the 35 references it cites were found in a registry or an index; 6 of the 9 sources its claims cite were checked for support.

Abstract

This study proposes a battery-aware transformer network (BATNet) for accurate and robust state-of-charge estimation in lithium-ion batteries for electric vehicles (EVs). BATNet combines feedforward neural networks with an encoder-only transformer enhanced by local attention to focus on recent signal trends and an adaptive forgetting mechanism to suppress transient anomalies. Simulations were conducted on a high-fidelity Hyundai KONA EV model in MATLAB Simulink, encompassing four driving cycles and five ambient temperatures. Results show that BATNet significantly outperforms traditional models, including gated recurrent units, recurrent neural networks, and long short-term memory networks, achieving a root mean square error as low as 0.0436%. An ablation study further verifies the contributions of the proposed architectural components. BATNet effectively captures nonlinear battery dynamics and remains stable under varying thermal and load conditions, making it well-suited for integration into real-time battery management systems in EV applications.

Overview

CoChat Summary

TL;DR

The paper proposes BATNet, an encoder-only transformer with feed-forward networks, local attention, and adaptive forgetting for state-of-charge (SOC) estimation in lithium-ion EV batteries. Using a simulated Hyundai KONA EV dataset across four drive cycles and five ambient temperatures, BATNet reports lower error than KF, GRU, RNN, and LSTM, with RMSE 0.0436% and R² 0.9996.

Key findings

  • BATNet achieved R² 0.9996, MAE 0.0349, MSE 0.0019, and RMSE 0.0436 on the KONA EV database, outperforming KF (R² 0.8887, MAE 0.3098, MSE 0.1467, RMSE 0.3831), 2Layer-GRU (0.9989, 0.0518, 0.0041, 0.0640), 2Layer-RNN (0.9991, 0.0467, 0.0031, 0.0557), and 2Layer-LSTM (0.9994, 0.0398, 0.0024, 0.0490).
  • Ablation results: full BATNet had R² 0.9995, RMSE 0.0063, MAE 0.0048, and 6 s time; removing local attention gave R² 0.9977, RMSE 0.0135, MAE 0.0104, and 15 s; removing adaptive forgetting gave R² 0.9991, RMSE 0.0083, MAE 0.0058.
  • BATNet consistently achieved the lowest RMSE across the full temperature spectrum; low temperatures (e.g., 273 K) increased RMSE due to higher internal resistance, reduced reaction kinetics, and transient current fluctuations. Removing local attention degraded low-temperature performance, while removing adaptive forgetting reduced performance under fluctuating load and cold-start conditions.
  • The dataset contained 10,454,039 total data points at 0.5 s sampling across five ambient temperatures (20°C, 10°C, 0°C, −10°C, −20°C) and four drive cycles (FTP75, HWFET, LA92, UDDS).

Methods

  • A high-fidelity Hyundai KONA EV model was built in MATLAB Simulink with driving, BMS, and battery/electrical subsystems; the battery module used a Thevenin equivalent circuit model. Simulations covered five ambient temperatures and four driving cycles, producing 10,454,039 data points at 0.5 s sampling. Preprocessing included time alignment, segmentation across drive cycles, moving-average smoothing, and three-sigma outlier removal.
  • BATNet combines feed-forward networks with an encoder-only transformer. The input shape is (batch_size, 287, 3); the architecture includes embedding, positional encoding, multi-head attention, adaptive attention with a learnable scalar/gating coefficient and modulation matrix, head-wise adaptive modulation, global pooling, and a regression head. Ablation variants removed local attention or adaptive forgetting.
  • BATNet was compared with KF, 2Layer-GRU, 2Layer-RNN, and 2Layer-LSTM; all models used a two-layer structure with matched complexity. Evaluation metrics were R², MAE, MSE, and RMSE.

Limitations

  • Stated: the current work assumes a fresh-cell scenario without incorporating degradation effects; direct evaluation on real vehicle datasets remains a future direction; the study relies on high-fidelity simulations using the Hyundai KONA EV model; edge-device deployment requires future model compression such as quantization or knowledge distillation; empirical validation on multi-chemistry datasets is planned.
  • Evident from the text: evaluation is simulation-based rather than real-world vehicle data, and no aging or multi-chemistry empirical validation is included. The text also reports different RMSE values in different tables (0.0436 in Table 5 vs. 0.0063 in the ablation table) without fully reconciling the evaluation settings.

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

Integrity signals

Registry

Crossref · 10.1093/jcde/qwaf042

Cites

35 of 35 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

4 Valid · 1 Not valid · 2 Not checked · 92 Own results and interpretations

38 more claims

Create a free account to see all 99 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 35
  • Probably not a substantial claim. 28

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.

3 of 3

Citations

Cites

Show all 35 references

Cited by

See all 5

Metadata

Authors
Mu, B.-Q., Doukhi, O., Wang, L., & Lee, D.-J.
Affiliations
Jeonbuk National University (KR)
Corresponding author
Lee, Deok-Jin
Venue
Journal of Computational Design and Engineering
Published
2025-06
Volume
12
Issue
6
Pages
73-95
ISSN
  • 2288-4300 (ISSN-L)
  • 2288-5048
OpenAlex
W4411646434
Cited by
5
References
34
Citations by year
2026: 3 · 2025: 2
FWCI
1.22
Type
journal article
Language
en
Open access
Locations
Journal of Computational Design and Engineering (publishedVersion, cc-by-nc)
Funders
  • Korea Institute for Advancement of Technology (R&D, P0025272, P0025300)
  • MOTIE (20018904)
Topics
  • Physical Sciences › Engineering › Automotive Engineering › Battery Modeling and Management
  • Physical Sciences › Engineering › Automotive Engineering › Electric and Hybrid Vehicle Technologies
  • Physical Sciences › Engineering › Electrical and Electronic Engineering › Electric Vehicles and Infrastructure
Keywords
state of charge estimation, lithium-ion batteries, electric vehicles, encoder-only transformer, local attention, battery management systems, battery dynamics, temperature effects, driving cycles

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: State-of-charge estimation of commercial electric vehicles by using battery-aware transformer network algorithms. https://app.cochat.ai/papers/10.1093/jcde/qwaf042
The paper
Mu, B.-Q., Doukhi, O., Wang, L., & Lee, D.-J. (2025). State-of-charge estimation of commercial electric vehicles by using battery-aware transformer network algorithms. Journal of Computational Design and Engineering. https://doi.org/10.1093/jcde/qwaf042

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.