【作者】 张凯；

【导师】 马建；

【作者基本信息】 长安大学 ， 车辆工程， 2020， 博士

【摘要】 电动汽车作为新能源汽车的重要组成部分,因其运行过程中无污染、噪声低、能效高等特点,受到了政府、企业和高校等部门越来越多的关注、投入和推广。而车载电源作为电动汽车的关键部件,如何增强动力电池的运行安全性、延长使用寿命,成为了研究热点和难点。复合电源,由动力电池和超级电容组成,能够利用超级电容充放电功率高的特点,降低动力电池的峰值功率,抑制动力电池的大电流充放电,达到延长动力电池使用寿命的效果,成为了一个重要研究方向。本文从状态估计和能量管理策略方面对复合电源展开研究,提出了相应的状态估计方法和能量管理控制策略,并基于A&D5435开发了复合电源硬件在环半实物实验平台,通过仿真分析和实验验证相结合的方法验证了所提出的状态估计方法和能量管理策略的控制效果。本文主要研究内容如下:(1)本文建立了电动汽车动力学模型、动力电池仿真模型、超级电容仿真模型;构建了半主动式复合电源控制结构。该结构包含多种控制模式,可以充分发挥超级电容的功率特点,降低使用成本,提高能量使用效率;针对复合电源模型参数辨识问题,本文基于蚁狮优化算法,结合混沌映射理论,开发了改进的蚁狮优化算法,进行复合电源的模型参数辨识。辨识结果表明,所提出的参数辨识算法收敛速度快,鲁棒性好,准确度高,能够得到较小的估计误差;基于连续功率能量函数方法,分析了典型工况下,超级电容需求容量与电池功率阈值的关系。根据车辆需求功率的分布,确定电池功率阈值,进而得到了超级电容参数配置。(2)针对能量状态估计问题,本文系统地分析了目前动力电池能量状态估计方法存在的问题与不足,结合容积卡尔曼滤波CKF和粒子滤波PF,开发了自适应权重粒子滤波AWCPF算法。首先,基于状态向量和测量向量的残差向量,提出了自适应权重容积卡尔曼滤波算法。其次,为了抑制粒子滤波PF的粒子衰退现象,同时也为了提高能量状态的估计精度,利用自适应权重容积卡尔曼滤波得到粒子滤波PF的近似最优建议分布。然后,粒子滤波PF算法根据得到近似最优建议分布,对动力电池的能量状态粒子进行重要性采样,得到能量状态的估计结果。最后,在多种工况下进行了仿真试验。仿真结果表明,相比于PF和CKF方法,AWCPF方法精度更高和鲁棒性更强,且收敛性分析表明,AWCPF方法能够快速收敛到的状态真实值。(3)针对电池健康状态估计问题,基于电池恒流恒压充电后的电压恢复曲线,提取了两种电池健康状态特征参数:样本熵和指数系数。然后基于支持向量回归机SVR算法,以样本熵和指数系数为输入,电池可用容量为输出,建立了电池健康状态估计模型。通过实验验证和对比分析,验证了基于SVR健康状态估计模型的有效性,表明了样本熵和指数系数作为电池健康状态特征参数,并用来进行电池健康状态估计的有效性。(4)针对复合电源控制策略问题,本文分析了半主动式复合电源的电流电压的非线性特征及模型参数的时变性,提出了基于自适应模型预测控制算法的能量管理控制策略。基于历史数据,采用本文提出的改进的蚁狮优化算法,对电池模型参数进行在线辨识,自适应的实时更新模型预测控制中的参数矩阵。采用自适应权重容积粒子滤波算法进行模型状态估计。在混合驾驶工况和真是西安行驶工况下进行实验验证,结果表明,自适应模型预测控制算法能够获得较好的控制效果,且能更好的适用于非线性时变模型。(5)针对复合电源控制策略实验验证问题,通过动力电池组、超级电容组、DC/DC变换器、新威电池测试系统和AD5435搭建了复合电源半实物仿真实验平台。在Matlab/Simulink环境中对控制模型和优化算法进行建模、调试,然后使用RTW模块自动生成C代码,下载到A&D5435中进行运算执行。实验结果表明开发的能量管理策略算法基本能够实现超级电容的“削峰填谷”的效果,体现复合电源系统的在减缓动力电池性能衰退,延长动力电池使用寿命上的优势。更多还原

【Abstract】 As an important part of new energy vehicles,electric vehicles have been paid more and more attention,invested and promoted by the government,enterprises,universities and other departments because of their non pollution,low noise and high energy efficiency.As the key component of electric vehicle,how to enhance the safety and prolong the service life of power battery has become a research hotspot and difficulty.Composite power supply,which is composed of power battery and super capacitor,can take advantage of the high charging and discharging power of super capacitor,reduce the peak power of power battery,restrain the high current charging and discharging of power battery,and achieve the effect of prolonging the service life of power battery,which has become an important research direction.In this paper,the state estimation and energy management strategies are studied,and the corresponding state estimation methods and energy management control strategies are proposed.Based on A&D5435,the hardware in loop experiment platform of the composite power system is developed.The control effect of the state estimation method and energy management strategy is verified by the combination of simulation analysis and experimental verification.The detailed contents are as follows:(1)In this paper,the dynamic model of electric vehicle,the simulation model of power battery and the simulation model of super capacitor are established,and the control structure of semi-active composite power supply is constructed.This structure contains many control modes,which can give full play to the power characteristics of the super capacitor,reduce the cost of use and improve the efficiency of energy use.Aiming at the problem of parameter identification of composite power supply model,based on ant lion optimization algorithm and chaos mapping theory,this paper develops an improved ant lion optimization algorithm to identify the model parameters of composite power supply.The identification results show that the proposed algorithm has fast convergence speed,good robustness,high accuracy,and can get small estimation error.Based on the continuous power energy function method,the relationship between the demand capacity of super capacitor and the power threshold of battery under typical conditions is analyzed.According to the distribution of vehicle demand power,the battery power threshold is determined,and then the super capacitor parameter configuration is obtained.(2)Aiming at the problem of energy state estimation,this paper systematically analyzes and discusses the existing problems and shortcomings of current power battery energy state estimation methods.Based on the residual vectors of state vector and measurement vector,an adaptive weighted Cubature Kalman filter algorithm is proposed.Combining PF and CKF,the adaptive weighted Cubature Particle filter(AWCPF)is develop to estimate the energy state of power battery.The Cubature Kalman filter is used to get the approximate optimal proposal distribution,which is then transferred to PF for importance sampling of the energy state particles,which can not only suppress the particle decay phenomenon of PF,but also improve the estimation accuracy of the energy state.Finally,the simulation results verify that AWCPF algorithm has higher accuracy and better robustness compared with PF and CKF method,and the convergence analysis shows that AWCPF method can quickly converge to the real state value.(3)Aiming at the problem of battery health state estimation,based on the voltage recovery curve after constant current and constant voltage charging,two characteristic parameters of battery health state are extracted: sample entropy and exponential coefficient.Then,based on support vector regression(SVR)algorithm,a battery health state estimation model is established with sample entropy and index coefficient as input and battery available capacity as output.Through the experimental verification and comparative analysis,the validity of the model based on SVR is verified.It shows that the sample entropy and exponential coefficient,as the characteristic parameters of battery health state are effective for battery health state estimation.(4)Aiming at the control strategy of hybrid power supply,this paper analyzes the nonlinear characteristics of current and voltage of semi-active composite power supply and the time-varying of model parameters,and proposes an energy management control strategy based on adaptive model predictive control algorithm.Based on the historical data,the improved ant lion optimization algorithm proposed in this paper is used to identify the parameters of battery model on-line,and update the parameter matrix of model predictive control in real-time.The adaptive weight cubature particle filter is used to estimate the model state.The results show that the adaptive model predictive control algorithm can obtain better control effect,and can be better applied to nonlinear time-varying model.(5)Aiming at the experimental verification of the control strategy of the composite energy system,a hardware in loop simulation experimental platform is built by using the power battery pack,super capacitor bank,DC / DC converter,Neware battery test system and A&D5435.The control model and optimization algorithm are modeled and tested in Matlab / Simulink environment,and then the C code is automatically generated by RTW module and downloaded to A&D5435 for operation.The experimental results show that the developed energy management strategy algorithm can basically achieve the effect of "peak shaving and valley filling" of supercapacitor,which reflects the advantages of composite enery system in slowing down the performance decline of power battery and extending the service life of power battery.更多还原