Estimating the state of charge (SOC) of compound energy storage devices in the hybrid energy storage system (HESS) of electric vehicles (EVs) is vital in improving the performance of the EV. The complex and variable charging and discharging current of EVs makes an accurate SOC estimation a challenge. This paper proposes a novel deep learning-based SOC estimation method for lithium-ion battery-supercapacitor HESS EV based on the nonlinear autoregressive with exogenous inputs neural network (NARXNN). The NARXNN is utilized to capture and overcome the complex nonlinear behaviors of lithium-ion batteries and supercapacitors in EVs. The results show that the proposed method improved the SOC estimation accuracy by 91.5% on average with error values below 0.1% and reduced consumption time by 11.4%. Hence validating both the effectiveness and robustness of the proposed method.

转移学习是开发性能RL代理的越来越普遍的方法。但是，尚不清楚如何定义源和目标任务之间的关系，以及这种关系如何有助于成功转移。我们提出了一种称为两个MDP或SS2的结构相似性的算法，该算法基于先前开发的双仿真指标来计算两个有限MDP的状态的状态相似性度量，并表明该量度满足距离度量的属性。然后，通过GRIDWORLD导航任务的经验结果，我们提供了证据表明，距离度量可用于改善Q学习剂的转移性能，而不是先前的实现。