【作者】 张坤；

【导师】 赵岭；

【作者基本信息】 聊城大学 ， 车辆工程（专业学位）， 2022， 硕士

【摘要】 “双碳”目标下,节能减排的力度进一步加大,纯电动汽车产业的崛起已成必然。然而“里程焦虑”问题始终是限制纯电动汽车发展难以避免的难题。本文以多挡AMT纯电动汽车为研究对象,提出了一种基于强化学习理论的智能换挡规律,同时,设计了一款高性能的变速器控制单元(Transmission Control Unit,TCU),以实现智能换挡规律的植入。首先,根据整车参数,基于Matlab/Simulink软件构建整车关键系统部件模型,即驾驶员模型,驱动电机模型,动力电池模型,传动系模型和车辆行驶动力学模型。然后,根据电机效率MAP图,设计了基于车速—油门踏板开度的两参数经济性换挡规律,并基于现有经济性换挡规律存在的问题,提出了一种基于强化学习理论的智能换挡规律。首先,考虑到强化学习“维度灾难”的问题,通过引入最优拉丁超立方设计(Optimal Latin hypercube design,Opt LHD)对以当前挡位、车速和加速度为状态变量的状态空间进行缩减,以降低控制算法算力;然后,以经济性换挡为目标,在兼顾换挡频率的条件下,根据不同挡位下的电机效率和SOC值设计回报函数,制定奖惩机制;最后,通过ε-greedy策略进行动作选择,并根据Q值的更新原则将该状态和动作下的Q值保存到Q表中,完成对智能换挡规律的设计。模型仿真结果表明,智能换挡规律可有效降低整车能耗,具有较好的自适应性。其次,在考虑TCU算力的条件下,通过分析TCU的功能需求,对主控芯片进行了选型,并对TCU各个模块的硬件电路进行设计,主要包括稳压电源供电电路、信号采集与处理电路、PWM电路和CAN通讯电路。此外,还针对TCU的系统架构,进行软件开发,包括信号采集程序开发,CAN通讯程序开发,换挡过程控制程序开发和AMT换挡主程序开发。最后,将TCU应用于纯电动汽车整车性能综合测试平台,构建了硬件在环试验平台。通过静态换挡试验,验证了TCU硬件的可行性。同时,将两种换挡规律分别植入到控制器硬件中,通过动态换挡试验进行能耗对比,验证了智能换挡规律存在的优势和应用潜力。更多还原

【Abstract】 Under the goal of "carbon peak and neutrality",the intensity of energy conservation and emission reduction has been further increased,and the rise of the pure electric vehicle industry has become inevitable.However,the problem of "range anxiety" has always been an unavoidable problem to limit the development of pure electric vehicles.Taking the Multigear AMT pure electric vehicle as the research object,this paper proposes an intelligent shift schedule based on reinforcement learning theory,at the same time,a high-performance transmission control unit(TCU)is designed to realize the intelligent shift schedule.First of all,according to the vehicle parameters,based on Matlab/Simulink software,the model of the key system components of the vehicle is constructed,namely the driver model,the drive motor model,the power battery model,the transmission train model and the vehicle driving dynamics model.Then,according to the motor efficiency MAP diagram,the two-parameter economic shift schedule based on the speed-accelerator pedal opening degree is designed,and based on the existing economic shift schedule,an intelligent shift schedule based on reinforcement learning theory is proposed.First of all,considering the problem of reinforcement learning "dimensional disaster",the optimal Latin hypercube design(Opt LHD)is introduced to reduce the state space with the current gear,vehicle speed and acceleration as the state variables to reduce the computing power of the control algorithm;then,with the goal of economic shifting,under the condition of taking into account the frequency of shifting,the return function is designed according to the motor efficiency and SOC value in different gears and a reward and punishment mechanism is formulated;Finally,the action selection is carried out through the ε-greedy strategy,and the Q value under the state and the action is saved to the Q table according to the principle of updating the Q value,and the design of the intelligent shift schedule is completed.Model simulation results show that the intelligent shift schedule can effectively reduce the energy consumption of the whole vehicle and have good adaptability.Secondly,under the condition of considering the computing power of the TCU,by analyzing the functional requirements of the TCU,the main control chip is selected,and the hardware circuit of each module of the TCU is designed,mainly including the voltage regulated power supply circuit,signal acquisition and processing circuit,PWM circuit and CAN communication circuit.In addition,software development is also carried out for the system architecture of the TCU,including signal acquisition program development,CAN communication program development,shift process control program development and AMT shift master program development.Finally,the TCU is applied to the comprehensive test platform for the performance of pure electric vehicles,and a hardware-in-the-loop test platform is constructed.Through static shift tests,the feasibility of TCU hardware is verified.At the same time,the two shift schedules are implanted into the controller hardware,and the energy consumption is compared through the dynamic shift test,which verifies the advantages and application potential of the intelligent shift schedule.更多还原