Browse by author
Lookup NU author(s): Professor Volker PickertORCiD
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
The integration of reconfigurable battery storage(RBS) systems into dc shipboard microgrids (SMG) holds con-siderable potential by improving cell-level flexibility and systemadaptability compared to conventional battery storage systems.This paper proposes an online-trained deep reinforcement learn-ing control (DRL) framework for an RBS system interfacedwith a bidirectional dual-active-bridge (DAB) converter, aimingto manage its dynamic operation while enabling adaptive currentregulation and cell balancing. Considering the nonlinear dynam-ics and real-time uncertainties inherent in practical SMGs, anautomated online training framework is developed to bridge thesim-to-real gap that often hinders the generalization of offline-trained DRL methods. To this end, the proposed online trainingframework is implemented on a dSPACE real-time platform,enabling the controller to continuously interact with the physicalsystem and refine its policy with real-world experiences. Exper-imental results validate that the proposed online-trained DRLapproach achieves convergence 20.1% faster, in terms of trainingepisodes, compared to simulation-based training. Comparativestudies demonstrate that the proposed control strategy achievessuperior dynamic performance, reducing settling time by 43.6%compared to the simulation-trained controller and by 53.7%compared to the benchmark predictive controller. It also enablesfaster charge-discharge transitions, completing mode switcheswithin just 9.1 ms. Furthermore, the approach improves batterybalancing and transient current handling, thereby improving itssuitability for SMG applications
Author(s): Yuan Z, Zeng Y, Ghias A, Pou Josep, Pickert V
Publication type: Article
Publication status: Published
Journal: IEEE Transactions on Transportation Electrification
Year: 2026
Pages: epub ahead of print
Online publication date: 12/01/2026
Acceptance date: 03/01/2026
ISSN (electronic): 2332-7782
Publisher: IEEE
URL: https://doi.org/10.1109/TTE.2026.3653172
DOI: 10.1109/TTE.2026.3653172
Altmetrics provided by Altmetric
