电池变温度模型似然函数参数辨识及SOC估计

Battery variable temperature model parameter identification by likelihood estimation and SOC estimation

锂离子电池模型参数随温度和荷电状态(SOC)的变化而变化,特别是低温时模型参数变化较大,单值模型不准确。为了解决温度变化对SOC估计的影响,提高模型的准确度,引入温度修正因子,建立了变温度二阶RC等效电路模型。在模型参数辨识过程中,提出了最大似然函数法(MLE)获取模型参数。MLE算法简单,可以解决最小二乘算法因数据增多出现的数据饱和问题,收敛性质随着样本数目增加变好。通过混合动力汽车Loadcycle工况实验验证了提出的参数辨识方法的正确性和准确性。在温度变化下的条件下,基于变温度等效模型,采用自适应扩展卡尔曼滤波(AEKF)算法估计电池的SOC,并与单值模型下采用AEKF算法和无迹卡尔曼滤波(UKF)算法得到的SOC估计值进行对比。通过对比可以看出变温度模型下的AEKF算法得到的SOC估计精度较高,误差在2.2%以内。

The internal resistance of lithium-ion batteries changes with the temperatures and the state of charge(SOC), the differences are especially larger at temperatures below zero. Thus, the battery model with changeless parameters is not accurate. In order to solve the problem that temperature changes have an impact on SOC estimation and enhance the accuracy of the model, temperature correction factors are pulled in. A variable temperature model with two RC networks is established. A maximum likelihood estimation(MLE) algorithm is proposed to obtain parameters in the process of model parameter identification. The MLE is simple, it can solve the data saturation problem caused by the increased data in the least squares algorithm. The convergence property of MLE becomes better as the number of samples increases. Loadcycle experiments are conducted to validate the correctness and accuracy of the proposed parameters identification algorithm. Adaptive Kalman filtering(AEKF) method is adopted to calculate the battery SOC based on the variable temperature model in a temperature-changeable environment. The result is compared with that based on single value model using AEKF and unscented Kalman filter(UKF). It can be seen from the comparison that the accuracy of SOC estimation obtained by AEKF is high, the error never exceeds 2.2%.