Scholarlink

<p>The advancement of electric vehicle technology increasingly depends on effective power management of hybrid energy storage systems that combine fuel cells, batteries, and supercapacitors. Traditional rule-based and optimization-based control strategies often fail to fully exploit these energy sources, especially under rapidly changing load conditions. This paper presents a reinforcement learning-based power management strategy, specifically using a Q-learning framework, to address these challenges. The proposed method eliminates the need for detailed system modeling by enabling the agent to learn optimal power distribution policies through continuous interaction with the environment. It dynamically considers battery state-of-charge, supercapacitor voltage, and instantaneous load demand to determine the most energy-efficient control actions. The proposed controller is trained with a carefully formulated reward function aimed at reducing hydrogen consumption, minimizing battery degradation, and effectively using the super capacitor for transient load compensation. Simulation results demonstrate that the proposed method outperforms conventional and optimization-based strategies. This controller achieves the lowest hydrogen consumption of just 15.58&#xa0;g. Furthermore, it delivers the highest system efficiency of 99.67%. These outcomes highlight the capability of the proposed approach to achieve substantial fuel savings while maintaining battery and super capacitor operation within safe limits. The effectiveness and robustness of the proposed reinforcement-learning-based energy management strategy are further validated under standard HWFET, UDDS, and NYCC driving cycles, demonstrating consistent fuel-consumption reduction and adaptive power sharing across diverse operating conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Reinforcement Learning Based Power Management & Cost Minimization in Fuel Cell, Battery & Super Capacitor Driven Electric Vehicle

  • 期刊论文

  • 发表日期：

    2026年4月15日

查看全文
  • S. Jerril Gilda,
  • A. Jose Anand

摘要

The advancement of electric vehicle technology increasingly depends on effective power management of hybrid energy storage systems that combine fuel cells, batteries, and supercapacitors. Traditional rule-based and optimization-based control strategies often fail to fully exploit these energy sources, especially under rapidly changing load conditions. This paper presents a reinforcement learning-based power management strategy, specifically using a Q-learning framework, to address these challenges. The proposed method eliminates the need for detailed system modeling by enabling the agent to learn optimal power distribution policies through continuous interaction with the environment. It dynamically considers battery state-of-charge, supercapacitor voltage, and instantaneous load demand to determine the most energy-efficient control actions. The proposed controller is trained with a carefully formulated reward function aimed at reducing hydrogen consumption, minimizing battery degradation, and effectively using the super capacitor for transient load compensation. Simulation results demonstrate that the proposed method outperforms conventional and optimization-based strategies. This controller achieves the lowest hydrogen consumption of just 15.58 g. Furthermore, it delivers the highest system efficiency of 99.67%. These outcomes highlight the capability of the proposed approach to achieve substantial fuel savings while maintaining battery and super capacitor operation within safe limits. The effectiveness and robustness of the proposed reinforcement-learning-based energy management strategy are further validated under standard HWFET, UDDS, and NYCC driving cycles, demonstrating consistent fuel-consumption reduction and adaptive power sharing across diverse operating conditions.