摘要
为解决燃料电池汽车在跟车场景下的生态驾驶问题,提出一种分层优化的解决办法(SAC-DP)。通过深度强化学习解决上层跟车速度控制问题,将汽车行驶过程中的多个目标融合于奖励函数,使燃料电池汽车在保持跟车安全的前提下提升跟车效率和舒适性;通过动态规划解决燃料电池汽车下层能量管理问题,减少汽车氢气消耗量和燃料电池的损耗,提高经济性。仿真结果表明,相较于Krauss-DP策略和CACC-DP策略,所提出方法至少可使舒适性提高27.26%,安全性提高21.66%,跟车效率提高10.08%,氢气消耗量减少4.13%,燃料电池损耗减少54.45%。
With the advances of autonomous driving and hydrogen fuel cell technology,energy management strategies for fuel cell vehicles in internet-connected environments have garnered keen academic attention.To address the eco-driving challenges of fuel cell vehicles in car-following scenarios,this paper proposes a hierarchical optimization solution(SAC-DP).The upper-level car-following speed control is realized by employing deep reinforcement learning,integrating multiple objectives of the driving process into the reward function to enhance following efficiency and comfort while ensuring safety.The lower-level energy management is achieved by employing dynamic programming,aiming to reduce hydrogen consumption and fuel cell degradation,thus improving overall efficiency.Results from two simulation scenarios indicate ride comfort improves by≥27.26%,safety by≥21.66%,following efficiency by≥10.08%,hydrogen consumption decreased by≥4.13%,and fuel cell degradation reduced by≥54.45%compared to two other strategies(Krauss-DP and CACC-DP).
作者
张玉坤
霍为炜
龚国庆
罗通强
ZHANG Yukun;HUO Weiwei;GONG Guoqing;LUO Tongqiang(Mechanical and Electrical Engineering School,Beijing Information Science&Technology University,Beijing 100192,China;Collaborative Innovation Center for Electric Vehicles,Beijing 100192,China;BYD Auto Industry Company Limited,Shenzhen 518118,China)
出处
《重庆理工大学学报(自然科学)》
北大核心
2026年第2期10-18,共9页
Journal of Chongqing University of Technology:Natural Science
基金
国家自然科学基金面上项目(52077007)。
关键词
生态驾驶
分层优化
燃料电池汽车
深度强化学习
跟车场景
eco-driving
hierarchical optimization
fuel cell vehicles
deep reinforcement learning
car-following scenario
