改进无迹粒子滤波的电动汽车锂电池SOC估算

Estimation of lithium battery SOC based on improved unscented particle filtering

针对粒子滤波对电动汽车锂电池荷电状态(SOC)估算误差大的问题,在建立二阶RC等效电路模型并利用脉冲放电实现电池参数辨识的基础上,采用了改进的无迹粒子滤波(IUPF)算法。该算法利用无迹卡尔曼在粒子滤波中生成重要的概率密度函数,然后在重采样阶段通过设置粒子阈值选择最优粒子,并用正则化粒子滤波改善了粒子退化问题。分别在恒流放电状态和动态应力测试(DST)下对该算法进行验证,实验结果表明:锂电池SOC估算最大误差为1.86%,提高了锂电池SOC估算精度,为电动汽车锂电池管理系统准确在线估计提供有效依据。

Aiming at the problem of large estimation error of state of charge(SOC)of lithium battery of electric vehicle by particle filtering,on the basis of establishment of the second-order RC equivalent circuit model and the use of pulse discharge to achieve battery parameter identification,an improved unscented particle filtering(IUPF)algorithm is used.The algorithm uses unscented Kalman to generate important probability density functions in particle filtering.Select the optimal particle by setting the particle threshold in the resampling stage,and improve the particle degradation problem by regularized particle filtering.The algorithm is verified under constant current discharge state and dynamic stress test(DST).Experimental results show that SOC estimation maximun error of lithium battery is up to 1.86%,improve the precision of lithium battery SOC estimation,Provide an effective basis for accurate online estimation of electric vehicle lithium battery management system.