电动汽车动力电池组不确定性建模和状态多尺度估计方法研究

Battery technology is the bottleneck with electric vehicles. It is important, both in theory and practical application, to do research on the state estimation, which is essential to optimizing energy management, extending the life cycle, reducing cost, and safeguarding the safe application of batteries in electric vehicles. This proposal focuses on the fundamental state estimation problem of lithium-ion battery packs used in electric vehicles. First, this proposal intends to build an n-order general battery model, and uses the information entropy theory to solve the balance calculation problem among model complexity, accuracy and real-time performance. To improve the state estimation accuracy of battery cell, a data and model fusion based joint estimation approach for battery parameter and state has been developed, and the influence factors from time-varying characteristics of battery system, sampling interval, calculation step and parameter degradation path have been investigated accordingly. Second, the proposal uses the statistical theory to analyze the coupling mechanism of the inconsistent characteristics for battery cells before and after their grouping, and the cell sample has been selected accordingly. Then, a real-time reliability battery prediction model has been built based on the uncertainty optimization method. Third, the trajectory of electricity performance for battery packs has been revealed through analyzing their strong time-varying, nonlinear and non-uniform complex characteristic. Then, to achieve accurate online state estimation of battery packs, the multi-dimensional multi-scale joint state estimation approach has been proposed. The research results of this project would provide new theoretical foundation for the battery management system of electric vehicles, and finally improve the energy efficiency of battery packs, both theoretically and meaningfully.

动力电池是电动汽车的技术瓶颈，其状态估计关乎能量管理、循环寿命、使用成本和安全，研究具有理论意义和应用价值。针对电动汽车锂离子动力电池成组状态估计的基础问题，研究：(1)建立动力电池单体n阶通用模型，采用信息熵理论研究模型复杂度、精度与实时性的权衡问题，建立基于数据-模型融合的参数和状态联合估计算法，探究系统时变特性、采样间隔、计算步长和参数变化路径等对状态估计的影响规律；(2)采用统计学理论揭示动力电池成组前后的不一致特性耦合机理，筛选单体样本表征总体水平，应用面向可靠性和实时性要求的不确定性优化方法，建立动力电池组预测模型；(3)研究动力电池组的强时变非线性非均一等复杂特性，揭示动力电池成组的电性能变化规律，提出动力电池组多维多尺度状态联合估计方法，实现动力电池组状态的高精度在线估计。研究成果将为电动汽车动力电池管理系统奠定新的理论基础，对提高动力电池组的能效具有重要的理论指导意义。

发展以电驱动为特征的新能源汽车是国际共识和我国的国家战略，动力电池及系统管理是技术瓶颈。针对模型建不精、状态估不准、系统管不好等急需解决的关键问题，项目从单体到系统、从理论到应用逐级深入：. 首先应用分数阶理论建立动力电池交流阻抗频域模型、细化其动态极化电压特性解析，发明了动力电池N阶电气模型和频域模型融合的建模新方法，解决单一模型的建模针对性和模型精度约束性等问题。第二，针对动力电池参数和状态的强耦合且不同步时变等特性，提出了一种基于数据-模型融合的动力电池参数和状态多尺度协同估计方法，以宏观时间尺度估计动力电池容量、微观时间尺度估计荷电状态(SOC)，解决了动力电池容量和SOC的耦合估计难题。第三、针对动力电池成组方式与电能量利用率和寿命衰退特性耦合关系不明确的难题，建立了动力电池组先串后并和先并后串两种连接方式的数值分析模型，理论推导了最大化动力电池组容量的拓扑结构。第四、推导了分数阶模型在线应用的时序递推方程，解决了模型离散化过程中由于短记忆准则带来的有色噪声问题，实现了容量增量老化识别法的实车应用。提出了基于Box-Cox变换和蒙特卡罗模拟的RUL方法，利用Box-Cox变换构建可用容量与循环数之间的线性模型，进而外推预测动力电池容量衰减。第五、提出一种基于曲线特征量提取的电池筛选方法，应用不确定性量化方法提出了具有多时间尺度的“筛选单体模型”+“模型偏差补偿”的动力电池组不确定性建模和SOC估计方法。最后，搭建了基于xPC Target的动力电池在环仿真实验平台，验证了动力电池SOC和容量SOH估计算法的精确性和鲁棒性，SOC实时估计误差在2%以内。. 本项目研究内容属于电气、机械、电化学、控制等多学科交叉领域，研究成果为电动汽车动力电池管理奠定了新的理论基础，对提高动力电池管理系统的智能化水平和核心技术的自主可控具有重要现实意义和推动作用。