基于深度强化学习模型TD3优化和改进的电动汽车制动能量回收策略
CSTR:
作者:
作者单位:

作者简介:

通讯作者:

中图分类号:

U469.72;TP18

基金项目:

国家重点研发计划项目(2019YFE0122600);湖南省教育厅重点科研项目(22A0423);湖南省自然科学基金项目(2023JJ60267, 2022JJ50073).


Electric vehicle brake energy recovery strategy based on deep reinforcement learning model TD3 optimization and improvement
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    双电机驱动电动汽车能够实现前后独立转矩分配, 进而可以获得更高的能量回收效率. 针对双电机驱动电动汽车, 提出一种基于双延迟深度确定性策略梯度算法(TD3)进行改进优化的制动能量回收策略. 该策略能在保障制动安全性和舒适性的同时, 实现制动能量回收的最大化. 首先, 构建基于深度强化学习的能量回收决策框架, 并设计一个综合考虑能量回收效果、安全性和舒适性的奖励函数; 然后, 采用TD3算法求解该决策过程, 并提出改进的优先经验回放机制, 以加速策略的收敛速度; 最后, 引入平衡探索的噪声策略, 增强算法探索与利用的能力. 通过Matlab/Simulink平台验证, 所提出算法在满足制动安全性和舒适性的前提下, 能够更高效地分配制动力, 有效地提高制动能量回收效率.

    Abstract:

    Dual-motor-driven electric vehicles can realize independent torque distribution between front and rear, which in turn can obtain higher energy recovery efficiency. This paper proposes an improved and optimized braking energy recovery strategy based on the twin delayed deep deterministic (TD3) policy gradient algorithm for electric vehicles with dual motor drive. The strategy can maximize braking energy recovery while ensuring braking safety and comfort. First, an energy recovery decision-making framework based on deep reinforcement learning (DRL) is constructed, and a reward function that integrates the energy recovery effect, safety and comfort is designed. Then, the TD3 algorithm is used to solve this decision process, and an improved prioritized experience replay mechanism is proposed to accelerate the convergence speed of the strategy. Finally, the paper introduces a noise strategy for balanced exploration to enhance the algorithm's ability to explore and exploit. Validated by the Matlab/Simulink platform, the proposed algorithm is able to distribute the braking force more efficiently and effectively, improving the braking energy recovery efficiency under the premise of satisfying braking safety and comfort.

    参考文献
    相似文献
    引证文献
引用本文

彭自然,贺振宇,肖伸平,等.基于深度强化学习模型TD3优化和改进的电动汽车制动能量回收策略[J].控制与决策,2025,40(8):2361-2372

复制
相关视频

分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2024-09-10
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2025-07-11
  • 出版日期: 2025-08-20
文章二维码