基于短板电池辨识的锂离子电池模组状态在线估计理论与方法研究

The cell-to-cell variations in Lithium-ion battery pack is the greatest challenge to the accurate and fast estimation of state of charge and state of health for the battery pack. The existing pack state estimation methods are mostly based on the state estimation of all cells in the pack, leading to the huge modelling cost and computing cost, and then the difficulty in real application. To reduce the estimation cell number and the modelling cost, a new pack state estimation method which is on the basis of identification of two weakest cells in the pack is proposed here. First, the method of weakest cell identification will be developed based on the study of pack consistency assessment criteria. Then, the adaptive weakest cell state estimation method will be developed by incorporating the model migration strategy. After investigating the dynamic evolution of pack consistency, including the characteristic time scales and weakest cell transfer mechanisms, the pack state estimation method will be finally established based on the sequential implementation of dynamic weakest cell identification and weakest cell state estimation. The realization of this project will be greatly helpful to the Lithium-ion battery modelling and state estimation, and provide the powerful technical support for the safe and long life operation of Lithium-ion battery system.

锂离子电池模组一致性问题是制约电池模组荷电状态和健康状态快速、精准估计的主要原因。现有模组状态估计方法常以同步估计模组内所有单体电池状态为前提，存在建模成本高、在线计算量大等本质缺陷，实际应用困难。鉴此，本项目提出以先辨识两个短板电池为前提的模组状态估计方案，进而显著减少在线估计单体个数、降低建模成本。首先，研究模组一致性在线评价准则，开发基于数据驱动分析的短板电池快速、可靠辨识方法；其次，研究电池模型迁移技术，构建模型参数在线更新机制，开发自适应的短板电池精准状态估计方法；再次，研究模组一致性动态演化规律，认清模组一致性变化时间尺度特征以及短板电池转移机制，最终形成以“短板电池动态辨识-短板电池状态估计”序贯实施为基础的模组状态在线估计技术，实现全寿命周期内的电池模组状态精准估计。本项目的实施将丰富电池管理系统建模理论和方法，为锂离子电池系统安全、长寿命、按需运行提供技术支撑。

锂离子电池模组一致性问题是制约模组关键状态快速、精准估计的关键。本项目针对这一问题，围绕电池参数差异辨识、关键状态估计，开展了相应研究。在电池一致性评估方面，提出了基于相异性矩阵和多维标度算法的电池内部参数差异辨识方法，通过同步辨识单体间容量差异、内阻差异、荷电状态一致性，实现了模组内部短板电池识别；电池状态估计方法方面，提出了基于联合滚动时域估计策略的电池荷电状态估计方法，实现了参数失配、测量不确定条件下的精准电池荷电状态估计，提出了基于优化策略的多约束功率状态估计方法，实现了电池功率状态精准估计；以此为基础，采用拉丁超立方抽样，系统分析了模组不一致性对状态估计的影响，为实现模组状态在线估计提供支撑。