Abstract: The accurate estimation of the state of charge (SOC) of power batteries is of great significance for the safety of vehicle operation and energy management. This paper uses a second-order resistor-capacitor (RC) equivalent circuit model, uses the forgetting factor recursive least squares (FFRLS) method for online parameter identification, and uses extended Kalman filtering (EKF) for SOC estimation. The offline charging and discharging data under several working conditions of the battery are used as training data, and the back propagation neural network (BPNN) is used for training. The estimated error of the extended Kalman is used as the training output of the neural network to compensate for the error of the extended Kalman filter. FFRLS and EKF are used to ensure the real-time performance of the system, and BPNN is used to improve the robustness of the system. It is proved through simulation that this method can effectively improve the accuracy of SOC estimation.

Key words: online parameter identification; recursive least squares method; BPNN; EKF

摘要： 动力电池荷电状态（SOC）的准确估计对汽车行驶安全和能量管理有着重要的意义。 文章使用二阶电阻-电容（RC）等效电路模型，利用带遗忘因子的递推最小二乘（FFRLS）法 进行在线参数识别，用扩展卡尔曼滤波（EKF）进行 SOC 估算。将电池几种工况下充放电离 线数据作为训练数据，利用反向传播神经网络（BPNN）进行训练，将扩展卡尔曼的估算误差 作为神经网络的训练输出，补偿扩展卡尔曼滤波的误差。使用 FFRLS 与 EKF 保证系统的实 时性，使用 BPNN 提升系统的鲁棒性，通过仿真证明，此方法可以有效提高 SOC 估算的精度。

关键词: 在线参数识别；递推最小二乘法；BPNN；EKF