An intelligent cloud-based battery management framework integrating GRU state-of-health estimation and double DQN...

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Materials Science

Background:

• Increasing electric vehicle (EV) adoption necessitates advanced battery management systems (BMS).
• Aggressive charging practices degrade lithium-ion batteries, impacting reliability, thermal stability, and cost.
• Current BMS often lack sophisticated methods for predicting battery health and optimizing charging.

Purpose of the Study:

• To develop a hybrid, cloud-assisted framework for intelligent battery management in EVs.
• To enable health-aware fast charging by accurately predicting battery State-of-Health (SoH).
• To optimize charging control using a reinforcement learning approach that enhances stability and avoids overestimation.

Main Methods:

• Utilized a Gated Recurrent Unit (GRU) model for nonlinear and temporal State-of-Health (SoH) prediction.
• Implemented a Double Deep Q-Network (Double DQN) algorithm for charging optimization within a reinforcement learning loop.
• Designed a cloud-assisted architecture for scalable parameter updates and fleet monitoring.

Main Results:

• The proposed GRU-Double DQN framework significantly minimized capacity degradation compared to traditional methods.
• Enhanced thermal regulation and optimized charging efficiency were observed.
• The system demonstrated adaptive charging capabilities, balancing speed, safety, and battery durability.

Conclusions:

• The GRU-Double DQN framework provides a robust and scalable solution for next-generation intelligent BMS in EVs.
• Health-aware fast charging is achievable, extending battery lifespan and improving overall system performance.
• Cloud integration facilitates efficient fleet management and continuous system improvement.