Answers are grounded in this paper and cite their sources.
The state of charge (SOC) of electric vehicle (EV) batteries must be known accurately to operate them efficiently, reliably and for a long service life. Conventional estimators depend on detailed battery models or additional sensing hardware, which makes them slow to set up, costly and error-prone. We propose an SOC estimator that couples deep reinforcement learning (DRL) with spider swarm intelligence (SSI). A Deep Q-Network (DQN) learns the estimation policy from measured battery signals without an explicit battery model, while SSI tunes the balance between exploration and exploitation during training so that learning converges faster and predictions become more accurate. On the public NASA PCoE lithium-ion battery dataset, the hybrid estimator lowers the largest SOC error from 30% to 8% and achieves clearly lower mean absolute error (MAE) and root mean square error (RMSE) than conventional techniques. Adding SSI to DRL also shortens convergence and makes the estimates more robust and precise. The hybrid method is therefore a promising candidate for advanced EV battery management systems, where it can raise both reliability and efficiency.
No takes yet. Share an insight, caveat, or question.
Afzal et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: