多源信息与干扰耦合下的电池多维、多尺度容量估计与融合

Capacity is one of the most important time-variant parameters indicating battery performance. For a better application and management of the battery systems, an accurate online capacity estimation is necessary. Two challenges exist in the design of capacity estimator. One is, on the cell level, that highly coupled multi-source information and noises lead to an inaccurate capacity calculation, and the other is, on the module or pack level, that the cell inconsistency within a battery module/pack introduces a nonlinear characteristics mapping between cell and module/pack. Aimed at solving the problems caused by the first challenge and thereby enhancing the estimation accuracy of cell capacity, we will investigate the characteristic of the multi-source noises, then with the noise characterization, the total least squares (TLS) based capacity estimation algorithm will be developed. With fully considering the multi-source information, a multi-dimension capacity estimator will be further designed. Based on the multi-dimension estimator, a fusion of the multi-dimension capacity estimation results will be designed with adaptive Kalman filtering. Aimed at solving the problems caused by the second challenge and enhancing the estimation accuracy of module/pack capacity, we will first study an averaged module/pack capacity estimation based on the multi-dimension estimator. Further, the nonlinear characteristics mapping between cell and module/pack and its mathematical expression will be well investigated, which facilitates the design of the correction of capacity estimation based on adaptive neuro - fuzzy inference system (ANFIS). It is expected that, with the research of the project, a solution called multi-dimension and multi-scale capacity estimation and fusion considering highly coupled multi-source information and noises will be obtained, which finally improves the battery management technology and enhance the performance and safety of battery systems.

电池容量是表征电池性能的一个重要时变参数，为了更好地管理及使用电池，容量的准确在线估计至关重要。容量估计的两个难点是：①电池单体层面，多源信息与干扰强耦合导致容量计算误差较大；②电池模组层面，电池单体不一致导致单体特性与模组特性的非线性映射。针对第①难点，本项目拟通过对多源干扰进行特征化，提出总体最小二乘容量估计算法，并充分利用多源信息，进行多维容量估计设计，在此基础上，研究提出基于自适应卡尔曼滤波的多维容量估计融合，提高单体容量估计的精度；针对第②难点，本项目拟在多维容量估计及融合方法基础上，研究单体到模组尺度变化后，平均意义上的模组容量估计，并进一步研究单体特性与模组特性的多尺度非线性映射关系及数学表达，提出基于自适应神经模糊推理的容量估计修正，提高模组容量估计的精度。本项目研究形成多源信息与干扰耦合下的电池多维、多尺度容量估计与融合方案，完善电池管理技术，提高电池使用的性能和安全性。

电池容量是表征电池性能的一个重要时变参数，准确的电池容量估计对电池管理至关重要。本项目针对电池容量估计展开研究，在单体层面考虑多源信息与干扰耦合问题，在模组层面考虑单体不一致的影响，最终实现电池多维、多尺度容量估计与融合。针对电池电压测量噪声以及电流测量偏差，分别设计了偏差补偿最小二乘算法与两阶段最小二乘算法，实现电池模型参数的无偏、准确辨识，为电池状态估计提供精确的模型信息；建立了可变多时间尺度下的电池SOC与容量联合估计算法，重点考虑了SOC估计不确定性、电流积分噪声以及状态间不同的变化特性对电池容量估计的影响；研究了基于容量增量分析的容量估计方法，提出基于卡尔曼滤波算法的容量增量曲线快速获取方法，探究了电池老化状态、充电初始SOC和温度对容量增量曲线的影响，最终构建了考虑充电初始SOC的容量估计修正方法；基于上述容量估计值，提出一种全工况下电池多维容量自适应融合方法，推导出容量估计误差协方差并构建容量融合的状态空间方程，通过卡尔曼滤波实现了多维容量自适应融合；针对电池单体不一致问题，以融合算法中的电池容量作为参考，构建了包含SOC与容量差异的差异方程，针对状态、状态差异之间不同的变化特性，建立了具有多时间尺度的“多参考-差异”模型，最终实现电池模组SOC与容量联合估计。本项目研究可以提升动力电池容量估计精度，提高电池使用性能及安全性，更加合理、准确地评估电池组的健康状态，为新能源汽车发展助力。