Issue |
EPJ Web Conf.
Volume 330, 2025
The 5th International Conference on Electrical Sciences and Technologies in the Maghreb (CISTEM 2024)
|
|
---|---|---|
Article Number | 06006 | |
Number of page(s) | 7 | |
Section | Electric Vehicles and Hydrogen Technologies | |
DOI | https://doi.org/10.1051/epjconf/202533006006 | |
Published online | 30 June 2025 |
https://doi.org/10.1051/epjconf/202533006006
Estimation of Battery SoC using Reinforcement Learning
LSIB Laboratory, FST, Hassan II University of Casablanca, Mohammedia 28806, Morocco
* Corresponding author: idriss.mortabit@gmail.com
Published online: 30 June 2025
Battery management systems are critical for electric vehicles, as they directly impact performance, safety, and how long the battery lasts. One of the biggest challenges is accurately estimating the State of Charge (SoC) - essentially how “full” the battery is. Energy management in electric vehicles relies heavily on robust tracking of power reserve. Traditional monitoring strategies are unable to handle dynamic and volatile behaviour of energy storage devices and do not provide for diverse operational scenarios. Our work utilized a high-end adaptive learning approach called PPO, which learns from operational data directly without structural assumptions beforehand. The approach capitalizes on real-world transportation data (electric factors, distances, and timing parameters) in envisioning reserve measurement as a perpetual challenge to learn. We constructed a real-time simulation world in which an AI system engages with virtual power cells to come up with maximum evaluation strategies. The architecture contains two networks of computation: a primary estimator which calculates current reserves of energy as a function of situation inputs, and a feedback evaluator providing prediction accuracy ratings to enhance the performance of the main network. This learning is obtained through gradual interaction with the simulated environment. Experimental validation indicated that our adaptive approach demonstrates improved performance under various operational conditions and addresses significant limitations inherent in existing methods. Future directions include enhancing system resiliency, extensibility, and tractability. Our findings suggest that adaptive learning frameworks present a promising direction for the development of power management technologies for electric mobility solutions.
© The Authors, published by EDP Sciences, 2025
This is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.