Deep learning model tracks EV battery health with high precision

by Riko Seibo

Tokyo, Japan (SPX) Feb 16, 2026

With electric vehicles and grid storage expanding worldwide, engineers are looking for better ways to track how lithium ion batteries age under real driving and operating conditions.
A new study supported by Jilin University and China FAW Group reports a deep learning based method that monitors battery state of health with errors below 1 percent even when current and voltage vary in complex patterns.

The work appears in the journal ENGINEERING Energy and focuses on state of health, a metric that reflects how much usable capacity remains compared to a fresh cell.
Conventional approaches often assume steady operating conditions and can struggle when faced with non monotonic voltage curves, irregular charging profiles, or partial charge data, all of which are typical for vehicles in daily use.

The research team developed a model they call Parallel TCN Transformer with Attention Gated Fusion, or PTT AGF.
This architecture runs two analysis streams in parallel, using a Temporal Convolutional Network to learn short term local patterns in the data while a Transformer module captures long range temporal dependencies and broader aging trends.

To feed these networks, the method extracts four health related features from dynamic charge segments that strongly correlate with true state of health.
The authors report that the correlation coefficients between these engineered indicators and laboratory measured state of health values exceed 0.95, providing a compact yet information rich description of battery condition.

An attention gated fusion block then combines the outputs from the TCN and Transformer.
This mechanism assigns adaptive weights to each feature stream so the model can emphasize whichever patterns are most informative at a given point in the battery life cycle, while downplaying noise or less relevant signals.

The team validated PTT AGF on three benchmark datasets from MIT, CALCE and Oxford that cover different cell chemistries, capacities and cycling protocols.
Across these tests, the model produced root mean square errors below 1 percent in all operating scenarios, a margin that the authors say surpasses many existing recurrent and convolutional neural network based methods.

On the CALCE data, the reported error is about 0.44 percent, and on the MIT dataset the error is about 0.77 percent.
The model also maintained high accuracy when only partial segments of the charge curve were available, demonstrating robustness when data are incomplete or measurements are noisy.

Beyond raw accuracy, the researchers examined how the attention mechanism behaves as batteries age.
They found that the learned attention patterns align with known degradation mechanisms, suggesting that the model is not only predictive but also offers some interpretability about which parts of the signal reflect capacity loss and internal changes.

According to the team, this combination of feature engineering, parallel deep learning and attention driven fusion could support more reliable battery management systems in electric vehicles and energy storage systems.
Better state of health tracking can enable safer operation, more accurate range prediction and optimized charging strategies that extend battery lifetime and reduce costs for manufacturers and users.

Research Report: Parallel deep learning with attention-gated fusion for robust battery health monitoring under dynamic operating conditions