【作者】 汪玉洁；

【导师】 陈宗海；

【作者基本信息】 中国科学技术大学 ， 控制科学与工程， 2017， 博士

【摘要】 环境污染和能源危机是世界各国面对的两大难题,电动汽车作为一种零排放的交通工具是解决环境与能源问题的一个优选方案。针对动力电池系统当前国内外的研究热点,本文依托复杂系统的建模与分析手段以及现代滤波技术,开展动力锂电池的建模、状态估计和管理策略的研究,目的是建立一套涵盖动力系统建模、估计与控制的理论与方法体系,为储能系统的精细化、安全化、高效化管理提供借鉴。本论文的主要工作如下:(1)精确的电池模型有助于电池的行为描述与分析,也是电池状态估计的基础和前提。锂电池本身是一个复杂的电化学系统,具有很强的非线性和时变性,同时电池的参数易受外界环境等因素的影响,增加了电池在随机环境下建模的难度。基于对电池内部机理和外部行为的分析,本文提出了一种基于数据驱动的电池建模与参数辨识方法。(2)提出了基于多模型切换策略的电池荷电状态估计方法,通过引入改进解释结构模型,实现了四种典型电化学模型的切换,通过实验对算法的精度和实时性进行了验证;分析了温度对库伦效率的影响,建立了电池的容量保持率模型,提出了一种基于容量保持率模型的荷电状态估计方法,在实验验证中分析了漂移电流对估计精度的影响;针对传统卡尔曼滤波器方法在解决非线性、非高斯系统的状态变量估计问题时容易出现发散的问题,提出一种基于贝叶斯估计理论的电池荷电和能量状态估计方法,并通过动态工况下的实验对该方法的精度和鲁棒性进行了验证。(3)研究了电池被动均衡和主动均衡的电路拓扑和均衡控制策略,针对被动均衡的不足,提出了一种基于双向DC/DC的主动均衡电路。基于对电池模型和电池状态估计方法的研究,提出了基于荷电状态的电池主动均衡控制策略,通过实验对比了不同均衡策略的性能。(4)将锂电池的建模与状态估计方法拓展到锂电池/超级电容混合储能系统的建模与状态估计中。创造性地提出了基于双滤波器的参数与状态联合估计器,采用扩展卡尔曼滤波器和无迹卡尔曼滤波器分别对锂电池和超级电容的模型参数以及荷电状态进行更新和估计。为了降低模型参数的收敛时间,在参数更新阶段采用递推最小二乘法寻找模型参数的初始最优值。通过比较锂电池和超级电容在动态放电工况下状态估计的精度,对本文提出的估计策略进行了验证。(5)将电池的建模、状态估计、均衡管理等技术集成到电池管理系统的应用中,可以提高电池管理系统的精细化、智能化管理程度,延长电池的使用寿命。通过分析管理系统积累的电池数据,可以进一步优化电池的建模、状态估计和管理策略。本文介绍了电池管理系统的软硬件架构,并对电动汽车电池管理系统进行了案例分析。更多还原

【Abstract】 Environmental pollution and energy crisis are two major challenges facing the world today.The electric vehicle as a kind of zero-emission transportations is a good solution for energy and environment issues.Aiming at the domestic and foreign research hotspots of the power battery system,this thesis studies the modeling,state estimation and management strategies of the power lithium-ion batteries based on modeling and analysis methods of complex system and modern filtering techniques.The main propose of this thesis is to build up a theory and method system which contains system modeling,estimation and control,and provide reference for fine,safe and efficient management of energy storage systems.The main work of this thesis can be summarized as follows:(1)An accurate battery model is beneficial to the description and analysis of battery behavior,which is also the basis and premise of battery state estimation.The lithium-ion battery is a complex electrochemical system with strong non-linear and time-varying characteristics,while the parameters of the battery are susceptible to external environment and other factors.Therefore,it is difficult for battery modeling in a random environment.Based on the analysis of the battery internal mechanism and external behavior,a data-driven battery modeling and parameter identification method is proposed in this thesis.(2)This thesis proposes a multi-model switching state-of-charge(SOC)estimation method.The improved interpretative structural modeling method is introduced to implement model switching of four typical electrochemistry battery models.The influence of temperature to the coulomb efficiency is analyzed in the thesis,and a battery capacity retention rate(CRR)model is established.A CRR model based SOC estimation approach is proposed and the influence of the drift noise to the estimation accuracy is analyzed in the validation test.In order to solve the divergence phenomenon which produced by the conventional Kalman filter method when observing the state variables in non-linear and non-Gaussian systems,a Bayesian estimation theory based SOC and state-of-energy(SOE)estimation approach is proposed.The experiments under dynamic working conditions are performed to verify the accuracy and robustness of the proposed method.(3)The circuit topology and control strategy of passive equalization and active equalization are studied.Aiming at the drawbacks of passive equalization,an active equalization circuit based on a bidirectional DC/DC is proposed.Based on the research of battery model and state estimation method,an active equalization control strategy based on SOC is proposed,and the performance of different equalization strategies is compared by experiments.(4)The modeling and state estimation methods of the lithium-ion battery can be extended to battery/ultracapacitor hybrid energy storage system.Therefore,a novel parameter and state co-estimator based on a dual-filter is proposed,where the extended Kalman filter and unscented Kalman filter are employed for parameter updating and SOC estimation of lithium-ion battery and ultracapacitor,respectively.To reduce the convergence time of the model parameters,the recursive least square algorithm is used to provide initial values with small deviation.By comparing the accuracy of state estimation of lithium-ion battery and ultracapacitor under dynamic discharge profile,the estimation strategy proposed in this thesis is verified.(5)By integrating the technologies include battery modeling,state estimation and balance management to the applications of battery management systems,the management degree of refinement and intelligence can be improved,and the battery life can be extended.The optimization of battery modeling,state estimation and management strategies can be implemented by analyzing the battery data accumulated in battery management systems.This thesis introduces the software and hardware frames of the battery management system.The case study of the battery management system for electric vehicles is analyzed.更多还原