基于EKF-LSTM算法融合的锂离子电池SOC估计

SOC Estimation of Lithium-ion BatteryBased on EKF-LSTM Algorithm Fusion

为了解决锂离子电池荷电状态(SOC)估计的实时性差、准确性低和累积误差大等挑战,此处提出了一种基于扩展卡尔曼滤波器(EKF)与长短期记忆(LSTM)神经网络相结合的融合算法。该算法以二阶RC等效电路模型为基础,对电池模型进行了参数辨识,基于EKF算法建立了SOC估计预测模型,并从减少数据突变以优化电池建模的角度引入LSTM神经网络,将锂离子电池的电流、电压、温度和更新的SOC值用作构建融合算法模型的输入。仿真和实验结果表明,该融合算法能够适应不同工况条件下的SOC估计,误差控制在1%以内,具有较高的估计精度和稳定性。

In order to address the challenges of poor real-time,low accuracy and high cumulative eror in lithium-ion battery state-of-charge(SOC)estimation,a fusion algorithm based on the combination of extended Kalman filter(EKF)and long-short-term memory(LSTM)neural network is proposed.The algorithm is based on a second-order RC equiv-alent circuit model,the battery model is parameter identified,a prediction model for SOC estimation is built based on the EKF algorithm,and the LSTM neural network is introduced from the point of view of reducing data mutations to optimize the battery modeling,and the currents,voltages,temperatures,and updated SOC values of the lithium-ion batteries are used as inputs to construct the model of the fusion algorithm.Simulation and experimental results show that the fusion algorithm is able to adapt to the SOC estimation under different operating conditions,and the error is controlled within 1%,with high estimation accuracy and stability.