Battery Management System for Edge Devices: Battery RUL Prediction and Charging Optimization Based on Incremental Learning and SAC-PSO

The efficient deployment of lithium battery management systems (BMSs) in edge computing environments is critical for extending battery life and improving charging efficiency. This study proposes a novel battery management method that combines incremental learning with the soft actor-critic (SAC)-particle swarm optimization (PSO) algorithm to enable real-time prediction of remaining useful life (RUL) of battery and optimization of current combinations for four-stage constant-current charging (4SCC). By integrating long short-term memory (LSTM) with incremental learning, the method adapts to newly added real-time data on edge devices without requiring the retraining of the entire model. Leveraging the SAC-PSO algorithm, dynamic charging strategies are adjusted based on predicted RUL and real-time data such as current and temperature rise, achieving efficient operation on the edge development board RK3588. Experimental results show that compared to the constant-current-constant-voltage (CC-CV) method, this approach reduces charging time by 11.51% while improving charging efficiency and battery lifespan by 2.20% and 23.76%, respectively. This study provides a novel solution for the application of lithium BMSs in edge computing environments.