Transformer-Enhanced Digital Twin Architecture for Robust and Interpretable Lithium-Ion Battery State-of-Health Prognostics
Архитектура цифрового двойника с усовершенствованием на основе трансформера для надежного и интерпретируемого прогнозирования состояния здоровья литий-ионных аккумуляторов
2026-07-01
SCID: 54.1/yzd42nup
Discuss with AI
Kalman filterdigital twinlithium-ion batteryphysics-guided learningstate-of-health estimation
Figures from the paper
Abstract (AI)
Accurate battery state-of-health (SOH) estimation is essential for reliable electric vehicle operation. This paper presents a unified digital twin framework that combines attention-based sequence modeling with physics-guided learning for robust SOH prediction. A Kalman filter-based refinement is incorporated to improve tracking stability under nonlinear degradation. The framework also includes visualization and uncertainty analysis to enhance interpretability. Experimental results on NASA lithium-ion battery datasets (B0033, B0034, and B0036) demonstrate improved prediction accuracy, stability, and cross-battery generalization compared to conventional methods. A real-time AI dashboard further validates its practical applicability for monitoring, remaining useful life estimation, and risk-aware decision-making.
Key Findings
1
A Kalman filter-based refinement improves SOH tracking stability under nonlinear battery degradation.
2
A real-time AI dashboard demonstrates practical applicability for battery monitoring and remaining useful life estimation.
3
A unified digital twin framework combines attention-based sequence modeling with physics-guided learning for lithium-ion battery SOH prediction.
4
Experiments on NASA datasets B0033, B0034, and B0036 show improved prediction accuracy, stability, and cross-battery generalization over conventional methods.
5
Visualization and uncertainty analysis enhance the interpretability of SOH prognostics and support risk-aware decisions.
Research Object
NASA lithium-ion batteries (datasets B0033, B0034, and B0036) undergoing degradation
Research Subject
robust, interpretable state-of-health prediction, tracking stability, uncertainty, and cross-battery generalization during degradation
Publication Details
Publication Date
2026-07-01
Journal
Publisher
ISSN
Cited by
0
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest