Abstract: Aiming at the problems of low accuracy of traditional Kalman filter algorithms in estimating SOC of lithium-ion batteries and poor adaptability to different temperatures, the fractional-order multi-innovations adaptive unscented Kalman filter (FOMIAUKF) algorithm is proposed to estimate the SOC of lithium-ion batteries. Firstly, adaptive genetic algorithms are used to identify the circuit model parameters under dynamic stress test conditions, and then a comparative experiment of SOC estimation of FOMIAUKF, FOUKF, and MIUKF algorithms is carried out in the U.S. federal city under operating conditions. Then, a comparison experiment of SOC estimation of FOMIAUKF, FOUKF, and MIUKF algorithms is carried out under the operating conditions of the U.S. federal city. The final results show that the FOMIAUKF algorithm has good adaptability at 0, 25, 45 ℃ temperatures, and the average absolute errors of the estimated SOC are 1.66%, 0.27% and 0.25%, respectively, and the root-mean-square errors are 1.72%, 0.41% and 0.39%, respectively, which are the lowest among the three algorithms, which is of great significance for the estimation of the SOC of lithium-ion batteries.

Key words: charge state estimation; fractional order model; adaptive genetic algorithm; multiinnovations adaptive unscented Kalman

摘要： 针对传统卡尔曼滤波算法估算锂电池荷电状态（SOC）精度不高，以及对不同温度适 应性较差的问题，提出了分数阶多新息自适应无迹卡尔曼滤波（FOMIAUKF）算法估计锂电 池 SOC。首先，采用自适应遗传算法在动态应力测试工况下辨识电路模型参数，然后在美国 联邦城市运行工况下对 FOMIAUKF、FOUKF、MIUKF 算法进行 SOC 估计对比实验。最终结 果表明，FOMIAUKF 算法在 0、25、45 ℃温度下具有较好的适应性，估计 SOC 的平均绝对 误差分别为 1.66%、0.27%、0.25%，均方根误差分别为 1.72%、0.41%、0.39%，在三种算法 中最低，对锂电池 SOC 估计具有重要意义。

关键词: 荷电状态估计；分数阶模型；自适应遗传算法；多新息自适应无迹卡尔曼