A Modular reconfigurable battery architecture with an adaptive real-time control strategy for optimal capacity utilization

Debba, G. B., Amamou, A., Kelouwani, S. et Sehli, N. (2026). A Modular reconfigurable battery architecture with an adaptive real-time control strategy for optimal capacity utilization. IEEE Open Journal of the Industrial Electronics Society . ISSN 2644-1284 DOI 10.1109/OJIES.2026.3652925

Prévisualisation

PDF Télécharger (7MB) | Prévisualisation

Résumé

Addressing the challenges of cell inconsistency in battery systems, a key factor affecting performance, dynamic reconfigurable batteries (DRBs) have emerged as a promising solution in recent research, ensuring better cell balancing and optimal energy use. However, existing designs often face trade-offs between control complexity, switching flexibility, and operational reliability, limiting their practicality for large-scale applications. To overcome these limitations, this study proposes a novel balancing strategy based on a modified modular architecture with flexible module-level switching. This design relies on a dual-layer control framework that simultaneously manages inter- and intra-module state of charge (SOC) equalization with reduced coordination overhead. A Genetic Algorithm-based approach is used for optimal switching control, enhanced by a Feedforward Neural Network (FNN) to predict SOC deviation, enabling efficient and reliable real-time balancing. Experimental validation on a lab-scale prototype and Hardware-in-the-Loop (HIL) simulations demonstrates the system's ability to improve battery capacity utilization and extend operating time by approximately 17% and 18%, respectively, compared with conventional system. Further testing on a large-scale pack with 320 cells demonstrates up to 40 minutes of additional autonomy over a fixed architecture, underlining the practical potential of the proposed method for real-world applications.

Type de document:	Article
Mots-clés libres:	Cell balancing Dynamic reconfigurable battery Feedforward Neural Network (FNN) Genetic algorithm Real-time control Vehicle autonomy
Date de dépôt:	26 févr. 2026 14:44
Dernière modification:	26 févr. 2026 14:44
Version du document déposé:	Post-print (version corrigée et acceptée)
URI:	https://depot-e.uqtr.ca/id/eprint/12638

Actions (administrateurs uniquement)

Éditer la notice