基于改进模型与优化自适应CKF的锂离子电池快速变温工况下的SOC估计

State-of-charge estimation of lithium-ion batteries in rapid temperature-varying environments based on improved battery model and optimized adaptive cubature Kalman filter

为实现锂离子电池在快速变温环境下高精度强鲁棒性的状态监测,本文提出了一种基于改进电池模型与优化自适应容积卡尔曼滤波器的锂离子电池荷电状态估计方法。首先,讨论了伪二维电化学模型与等效电路模型中对于电池荷电状态定义上的差异,并通过中间变量来修正传统等效电路模型中安时积分法计算得到的荷电状态结果,提出了一种新的改进电池模型。其次,基于多组恒温环境下所获得的锂离子电池开路电压测试数据与动态应力测试工况数据获取了所建立模型与环境温度相关的各项参数。同时,基于矩阵对角化原理与协方差矩阵自适应原理改进了传统的容积卡尔曼滤波器,进一步提升了整体算法的稳定性和处理随机采样噪声的能力。最后,在快速变温环境中6组不同的电池工况下验证了所建立改进电池模型的精度以及存在随机采样噪声干扰时所提荷电状态估计方法的有效性。结果显示,所提出的荷电状态估计方法适用于快速变温环境下的各类电池工况,在随机采样噪声干扰下估计结果的均方根误差均在1.3%以内。

In pursuit of high-precision and robust state monitoring of lithium-ion batteries under an environment with rapid temperature fluctuations,we propose a state-of-charge(SOC)estimation method based on an improved battery model and an optimized adaptive cubature Kalman filter(CKF).First,the discrepancies in SOC definition between a pseudotwo-dimensional electrochemical model and an equivalent circuit model are discussed.Introducing the improved battery model,the SOC results from the equivalent circuit model,calculated by ampere-hour integration,are rectified using intermediate variables.Subsequently,model parameters influenced by environmental temperature are identified from open-circuit voltage and dynamic stress test data under various constant-temperature environments.Moreover,the traditional CKF is optimized based on principles of matrix diagonalization and adaptive covariance matrix,bolstering overall stability and the ability of the proposed SOC estimation method to handle random sampling noise.Finally,experimental validation under six diverse battery operating conditions in rapidly temperature-varying environments demonstrates the accuracy of the established improved battery model and the effectiveness of the proposed SOC estimation method,even under random sampling noise.The results demonstrate the versatility of the proposed SOC estimation method across various battery operating conditions in rapidly temperature-varying environments,with an estimated root mean square error of approximately 1.3% under random sampling noise.