自适应扩展卡尔曼滤波电池荷电状态估算方法

Adaptive extended Kalman filter for estimating the charging state of battery

摘要:

为更精确预估电动汽车动力源的荷电状态,优化戴维南等效电路模型,用自适应扩展卡尔曼滤波进行荷电状态估算。对实验电池进行外特性数据获取实验,分别得到在充放电状态下的开路电压曲线,在开路电压-荷电状态对应曲线中考虑充放电状态变化的因素。对离线参数辨识进行优化处理,在固定参数离线辨识的基础上考虑充放电状态和荷电状态,并与在线辨识进行端电压估算对比。基于优化后电池模型,通过自适应扩展卡尔曼滤波及其对比算法估算荷电状态,在复杂脉冲电流工况下对比端电压和荷电状态的估算精度。结果表明:采用优化后电池模型离线辨识的端电压估算精度达到0.01 V以内,高于在线辨识的端电压估算精度。在优化模型及离线辨识基础上构建自适应扩展卡尔曼、扩展卡尔曼和交互多模型的算法模型,估算电池荷电状态。经动态工况实验验证,基于优化模型自适应算法的荷电状态估算精度达到0.05,高于交互多模型-扩展卡尔曼滤波算法及扩展卡尔曼滤波算法。

Abstract:

To more accurately estimate the state of charge of the power source of EVs, Thevenin equivalent circuit model is optimized, and the charge of state is estimated by adaptive extended Kalman filter. Firstly, the external characteristic data of the experimental battery and the open circuit voltage curves under charge and discharge state are obtained. The factors of charge and discharge state change are added to the corresponding curve of open circuit voltage-charge of state. Secondly, in the aspect of parameter identification, the off-line identification is optimized. The charge-discharge state and charge of state are considered on the basis of the off-line identification of fixed parameters. The estimation of terminal voltage is compared with on-line identification. Finally, based on the optimized battery model, the charge of state is estimated by adaptive extended Kalman filter and its comparison algorithm. And the estimation accuracy of terminal voltage and charge of state is compared under complex pulse current conditions. Experimental results show that the accuracy of terminal voltage estimation for off-line identification of optimized battery model is less than 0.01 V, which is higher than that for on-line identification. Based on the optimization model and off-line identification, adaptive extended Kalman and extended Kalman and interactive multi-model algorithm are constructed to estimate the charged state of the battery. The experimental results show that the estimation accuracy of charged state based on optimization model adaptive algorithm is 0.05, which is higher than that of the two contrast algorithms. The accuracy of adaptive extended Kalman filter based on optimization model is higher than that of interactive multi-model extended Kalman filter and extended Kalman filter.