面向性能在线评估的锂离子电池机理模型仿真关键技术研究

The multi-physics model of lithium ion batteries is time consuming and the parameters are difficult to obtain, which restrict its application in battery management systems. Analytical solutions of the internal mechanisms are acquired by means of mathematical derivation, numerical approximation and mesh grid reduction and replace the large-scale space iterative process. This solves the contradictions of multi-physics model in speed, accuracy and versatility. Frequency response analysis, and time domain steady-state/transient analysis are applied in designing of signal stimulus to decouple the interactions of various parameters in different internal processes. Then some of the separated parameters are determined through response analysis. Numerical method for parameter identification is time-consuming and easily to fall into local optimal solution. The sensitivity values of the remaining parameters are calculated. The relationship between sensitivity values and operating conditions is expressed in form of sensitivity matrix, according to which, the best identifiable operating condition of parameters in different sensitivity levels are designed and used in the identification process of a stepwise multi-objective genetic algorithm. The identification process is confined and supervised by the mechanisms of battery, which ensures the efficiency and reliability of the identification results. This project belongs to an applied fundamental research of inter-discipline involving electrical engineering, electrochemistry and applied mathematics, the research findings of which will be useful to the real-time evaluating of Li-ion batteries' performance and significant for modeling and evaluating other energy storage batteries.

锂离子电池机理模型存在数值求解复杂、参数难以获取的缺点，是制约其应用于电池管理系统的根本原因。本项目利用数学推导、数值近似、减少网格划分等方法获取机理模型内部复杂过程的解析解，从而替代数值求解的大规模空间迭代过程，解决机理模型仿真在速度、精度和通用性方面的矛盾；针对模型参数在电池内部过程中相互耦合，难以通过外部测试获取的问题，依据电池内部过程的时频域特性，设计不同时频域激励将内部过程解耦，并利用激励响应分析法获取机理模型的部分参数；针对数值方法辨识模型参数费时且容易陷入局部最优的问题，仿真得到敏感度与工况的关联矩阵，基于最佳辨识工况设计分步多目标遗传算法，用电池机理特点对辨识过程进行限定和指导，提高辨识方法的效率和可靠性。本项目属于电气工程与电化学、应用数学相交叉的应用基础研究。研究成果对基于机理模型的锂离子电池在线性能评估具有实用价值，对其它种类储能电池的模型仿真和性能评估具有参考意义。

锂离子电池机理模型存在数值求解复杂、参数难以获取的缺点，是制约其应用于电池管理系统的根本原因。本项目利用数学推导、数值近似方法获取机理模型内部复杂过程的解析解，在此基础上利用热阻模型描述电池产热和散热行为，解决机理模型仿真在速度、精度和通用性方面的矛盾；针对简化机理模型参数在电池内部过程中相互耦合，难以通过外部测试获取的问题，依据内部过程的时频域特性，设计不同时频域激励将内部过程解耦，并利用激励响应分析法获取模型参数；针对简化机理模型参数进行了敏感性分析，确定了在全寿命周期需要准确辨识的包括高敏感参数和低敏感参数在内的9个参数，并设计了一组充放电交替、倍率按一定幅度变化的改进辨识工况。基于上述研究内容获得了一系列丰富的具有实际应用价值的研究成果：实现了电池宽温度范围的高精度SOC估计，提出了一种基于简化机理模型的无损、快速低温预热方法，并且基于简化机理模型实现了电池老化机理分析与电池组筛选。