【作者】 王宇；

【导师】 王桐； 周成功；

【作者基本信息】 哈尔滨工程大学 ， 电子与通信工程（专业学位）， 2020， 硕士

【摘要】 随着生活水平的逐步提高,人们对环境的重视程度日益增加,在国家相关政策的大力推动下,电动汽车开始大规模普及。随之而来的高成本、续航里程不足和充电体验较差等问题也亟需解决。作为一种创新的商业化解决方案,电动汽车的换电模式逐渐成为研究热点,随之而来的换电基础设施短缺、利用率不均衡等现象也亟待解决,使换电模式能源补充快的优势得以充分发挥。本文借助掌握的相关知识,分析了行进中的电动汽车换电过程中存在的问题和制约换电等待时间的因素。利用智能交通系统中车辆与路侧单元、路侧单元与换电站之间的实时通信,针对行进中电动汽车换电调度问题进行分析。为提高换电站的利用率,减少电动汽车换电的等待时间,本文提出了一种基于强化学习的调度方案,并在该方案基础上进一步提出了一种双层可划分区域、面向智能体的换电调度策略。本文的研究主要包括以下两个方面:(1)提出了一种基于强化学习的电动汽车换电调度策略(Reinforcement Learning Based Electric Vehicle Battery Switch Strategy,RLBS):首先给出城市智能交通系统下的电动汽车换电调度环境模型,然后同时考虑换电站的利用率和电动汽车的换电等待时间两方面因素,并根据行进中电动汽车换电调度的移动性特点,设计了状态转移方案、奖励策略以及基于回溯最大奖励的价值函数更新机制。(2)在RLBS的基础上,进一步提出了一种双层可动态区域划分、基于强化学习的换电调度策略(Dynamic Area Division Battery Switch Strategy based on Reinforcement Learning,DABS):重新建立换电调度模型和区域划分模型,策略选择的目标从换电站转移到了换电区域,并根据模型重新设计了换电调度策略和区域划分策略进一步均衡换电站利用率和电动汽车的换电等待时间。最后,在ONE仿真平台上通过实验验证DABS不同参数取值对策略系统相关指标的影响,并通过与现有经典策略方案比较,对DABS进行评价。实验结果表明,本文所提出的DABS调度策略可以有效提高换电站的利用率,降低电动汽车换电的平均等待时间。更多还原

【Abstract】 With the gradual improvement of living standards,people focus more on the environment.With the implementation of related policies by the country,electric cars are widely spread.Consequently,the related problems such as high cost,insufficient endurance mileage and poor charging experience need to be solved urgently.As an innovative commercial solution,the electric vehicle switch mode is re-mentioned after several years of silence.Meanwhile,circumstances such as the shortage of battery switch infrastructure and uneven utilization rate need be resolved,which provides a good stage for electric switch mode with the advantage time efficiency.With the help of relevant knowledge,the existing problems and the factors that restrict the waiting time for electric vehicles during the electric vehicle battery switch,and used the realtime communication between vehicles and roadside units,roadside units and the substation in the intelligent transportation system are analyzed.The paper focuses on solving the problem of electric vehicle power transmission scheduling during traveling,improving of the utilization rate of the power station and reducing the waiting time for electric vehicle power conversion,with a scheduling scheme based on reinforcement learning proposed to determine the best strategy for this problem.Based on this scheme,a double-layer partitionable,intelligentoriented battery switch scheduling strategy is proposed.The research in this paper mainly includes the following two aspects:(1)A Reinforcement Learning Based Electric Vehicle Battery Switch Strategy(RLBS)is proposed: a system environment model for the urban intelligent transportation system is established,which fully considers the utilization rate of the power station and the waiting time for batteries switch.Based on the characteristics of the electric vehicle’s electric vehicle power dispatching in progress,a state transition scheme,reward strategy,and a value function update method that can review the maximum value are designed.(2)On the basis of RLBS,Dynamic Area Division Battery Switch Strategy based on Reinforcement Learning(DABS)is further proposed: in order to solve the high concurrency problem when an electric vehicle chooses a power station during traveling,and further improve the utilization rate of the substation in urban scenarios,the competition and cooperation between the substations are evaluated.A strategy for the RLBS battery switch area division is proposed to further balance the utilization rate of the substation and the waiting time for batteries switch.Finally,the effects of different parameters of DABS on the relevant indicators of the strategy system are discussed experimentally on the ONE simulation platform.The DABS is evaluated by comparing with the existing classic strategy schemes.The experimental results have shown that the DABS scheduling strategy proposed in this paper can effectively improve the utilization rate of the substation and reduce the average waiting time of battery switch for electric vehicles.更多还原