The recent increase in electric vehicles (EVs) has led to an escalation in the need for an intelligent battery management system that can deliver safe, efficient, and long-lasting lithium-ion battery operation. The aggressive charging practices result in battery degradation and exert a strong impact on the reliability, thermal stability, and lifecycle cost. The present paper suggests a hybrid framework using clouds to assist in the development of a State-of-Health (SoH) prediction using a Gated Recurrent Unit (GRU) with Double Deep Q-Network (Double DQN) charging optimisation to facilitate health-aware fast charging. The GRU model can capture both nonlinear and temporal degradation characteristics in historical battery characteristics, which allows real-time and correct estimation of SoH. These projections will form part of a double DQN reinforcement learning loop to eliminate the overestimation of Q-values and enhance policy stability in the process of charging control. A cloud-assisted architecture to update the parameters with scalable updates and monitor the fleet of vehicles also supports the proposed system. The performance of the simulation proves that the framework can minimise the capacity degradation, enhance thermal regulation, and optimise the efficiency of the charging process in contrast with the traditional constant-current and single-network DQN strategies. The co-discovery design facilitates making adaptive charging choices that trade off speed, safety, and long durability of the battery. All in all, the proposed GRU-Double DQN framework offers a strong and scalable framework to be used in the next-generation intelligent battery management system in electric vehicles.
Naresh et al. (Tue,) studied this question.