基于改进多新息扩展卡尔曼滤波的电池SOC估计

为了提高SOC估计精度，提出基于遗忘因子改进多新息扩展卡尔曼滤波（FMIEKF）方法. 建立锂离子电池的双极化等效电路模型，开展开路电压测试. 通过递归最小二乘法，实现电池模型参数在线辨识. 提出FMIEKF进行SOC估计，该方法在融合多新息辨识理论和卡尔曼滤波基础上，引入遗忘因子削弱历史数据修正权重，解决数据过饱和问题. 通过实验和硬件在环进行验证. 结果表明，FMIEKF具有较高的准确性和收敛性，最大估计误差为0.948%，平均误差为0.214%，在不同SOC初值下20 s内收敛，可以适用于实际的电池管理系统中.

Estimation of state of charge of battery based on improved multi-innovation extended Kalman filter

An improved multi-innovation extended Kalman filter was proposed based on the forgetting factor in order to improve the accuracy of SOC estimation. Dual-polarization equivalent circuit model of lithium-ion battery was established, and open-circuit voltage testing was conducted. Recursive least squares method was used to realize online battery model parameter identification. FMIEKF was proposed for SOC estimation based on the fusion of multi-innovation identification theory and Kalman filtering. A forgetting factor was introduced to weaken the historical data correction weight and solve the problem of data oversaturation. The method was verified through experiments and hardware-in-the-loop. The experimental results show that FMIEKF has higher accuracy and convergence. The maximum estimation error was 0.948%, the average error was 0.214%, and the FMIEKF converged within 20 seconds under different initial values of SOC. The method can be applied to the actual battery management system.