Technical gazette, Vol. 32 No. 5, 2025.
Original scientific paper
https://doi.org/10.17559/TV-20250118002279
Hybrid LSTM-Transformer and EPSO-EKF Framework for Advanced Battery Management Systems in Electric Vehicles
D. Manoj
orcid.org/0000-0003-4786-3607
; Department of Electrical and Electronics Engineering, S.S.M Institute of Engineering and Technology, Dindigul-624 622 Tamil Nadu, India
*
F.T. Josh
; Department of Electrical and Electronics Engineering, Karunya Institute of Technology and Sciences, Coimbatore, Tamil Nadu, India
* Corresponding author.
Abstract
The increasing adoption of Electric Vehicles (EVs) as a sustainable solution necessitates efficient and reliable Battery Management Systems (BMS) to mitigate greenhouse gas emissions. Lithium-ion batteries (LIBs), commonly used in EVs, face challenges in accurately estimating critical parameters such as State of Charge (SOC) and State of Health (SOH) due to complex battery dynamics and operational variability. SOC indicates the available energy relative to the battery's capacity, while SOH reflects the battery’s capacity to retain charge. Inaccurate parameter estimation can lead to overcharging, over-discharging, and safety risks. This paper presents a novel hybrid BMS framework that integrates Long Short-Term Memory (LSTM) networks, Transformer architectures, and the Enhanced Particle Swarm Optimization-Extended Kalman Filter (EPSO-EKF). The LSTM-Transformer model effectively captures temporal dependencies and attention mechanisms, enhancing SOC and SOH estimation. The EPSO-EKF dynamically adjusts noise covariance, ensuring robust performance under diverse operating conditions. Evaluation using standard driving cycles, such as US06 and FUDS, demonstrates significant improvements in estimation accuracy. For example, the LSTM-Transformer model achieved a 15% reduction in RMSE for SOC and a 12% reduction in RMSE for SOH compared to standard LSTM models, with Adjusted R² values of 0.98 for SOC and 0.95 for SOH. This paper also reviews intelligent control strategies and advanced algorithms for BMS design, highlighting their customization potential, efficiency, and limitations. The findings emphasize the potential of integrating AI and optimization techniques for next-generation BMS in EV applications, advancing sustainable transportation.
Keywords
artificial intelligence (AI); battery management system (BMS); electric vehicles (EVs); optimization algorithm; state of charge (SOC)
Hrčak ID:
335088
URI
Publication date:
30.8.2025.
Visits: 832 *