基于强化学习的混合动力运载工具能量管理策略综述  

A Review of Reinforcement Learning Based Energy Management Strategies for Hybrid Transportation Vehicles

作　　者：庞圣钊 赵巳雨 程博[1,2] 田文龙 毛昭勇[1,2,3] PANG Shengzhao;ZHAO Siyu;CHENG Bo;TIAN Wenlong;MAO Zhaoyong(Unmanned System Research Institute,Northwestern Polytechnical University,Xi’an710072,China;National Key Laboratory of Unmanned Aerial Vehicle Technology,Xi’an710072,China;School of Marine Science and Technology,Northwestern Polytechnical University,Xi’an710072,China)

机构地区：[1]西北工业大学无人系统技术研究院,西安710072 [2]无人飞行器技术全国重点实验室,西安710072 [3]西北工业大学航海学院,西安710072

出　　处：《无人系统技术》2025年第1期1-19,共19页Unmanned Systems Technology

基　　金：国家自然科学基金(52307252、U24B20115);秦创原引用高层次创新创业人才项目(QCYRCXM-2022-336)。

摘　　要：近年来,随着数据科学的不断发展,越来越多的研究尝试使用强化学习来解决多源混合动力无人运载工具能量管理问题。首先,结合国内外的研究现状,分析了地面、轨道、水路及航空运载工具的动力总成结构,并综述了使用强化学习进行能量管理的研究。其次,创新性地依据算法类型,将相关研究归为传统强化学习与改进强化学习两类,并针对状态-动作空间划分、超参数调节、奖励函数设计及算法适用性等关键维度展开探讨。然后,汇总分析运用强化学习解决运载工具能量管理问题的挑战。最后,对该领域未来发展予以展望,大模型、多模态学习、多传感器信息融合等技术,或许会为混合动力运载工具能量管理问题的优化与落地提供更多可能。In recent years,with the continuous development of data science,more and more researches try to use reinforcement learning to solve the energy management problem of multi-source hybrid unmanned vehicle.Firstly,combined with the research status at home and abroad,this paper analyzes the powertrain structure of ground,rail,waterway and air vehicles,and summarizes the research on energy management using reinforcement learning.Secondly,according to the type of algorithm,the related research is divided into traditional reinforcement learning and improved reinforcement learning,and the key dimensions such as state action space division,super parameter adjustment,reward function design and algorithm applicability are discussed.Then,the challenges of using reinforcement learning to solve the energy management problems of hybrid vehicles are summarized and analyzed.Finally,the future development of this field is prospected.Large model,multimodal learning,multi-sensor information fusion and other technologies may provide more possibilities for the optimization and implementation of energy management problems of hybrid vehicles.

关 键 词：能量管理 深度学习 强化学习 运载工具 混合动力系统 人工智能 交通电气化 

分 类 号：TM92[电气工程—电力电子与电力传动]