基于核极限学习机与容积卡尔曼滤波融合的锂电池荷电状态估计

Estimation on State of Charge of Lithium Battery Based on Fusion of Kernel Extreme Learning Machine and Cubature Kalman Filter

为了提高锂电池荷电状态(SOC)估计精度,在容积卡尔曼滤波(CKF)算法中引入误差补偿机制,提出一种核极限学习机(KELM)和CKF相融合的估计方法.通过递归最小二乘法(RLS)动态跟踪等效电路模型的参数,由CKF算法得到SOC的初步估计值;以锂电池的工作电压、电流、CKF算法的残差均值和方差作为输入,初步估计值与真实值的偏差作为输出,采用KELM算法,在联邦城市驾驶工况(FUDS)下进行训练,得到CKF算法估计误差的预测模型;利用KELM预测模型的回归预测能力对初步估计值进行误差补偿,从而达到降低估计误差的目的.为了验证所提方法的有效性和先进性,采用Arbin电池测试平台的实验数据进行仿真分析,结果表明在多种工况下所提方法均具有更高的估计精度、更强的泛化性和鲁棒性.

To improve the State of Charge(SOC) estimation accuracy,an improved Cubature Kalman Filter(CKF) algorithm is proposed,in which an error compensation module based on Kernel Extreme Learning Machine(KELM) algorithm is integrated into CKF properly.The Recursive Least-Square method(RLS) is used to online estimate the parameters of the equivalent circuit model of the battery,and the preliminary SOC value is estimated using the CKF algorithm accordingly.Then,taking the operation voltage and current of the battery,and residuals mean and variance of the preliminary SOC estimation values as the input,and the corresponding SOC estimation error as the output,the SOC estimation error prediction model based on the KELM algorithm is obtained,and trained by using the Federal Urban Driving Schedule(FUDS) condition data.The regression prediction ability of the KELM model is utilized to perform error compensation on preliminary estimation,so as to reduce the error of SOC estimation.To verify the effectiveness and advancement of the proposed method,the experimental data of the Arbin battery test platform are used for simulation analysis.The results show that the proposed method has higher estimation accuracy,stronger generalization,and robustness under various operating conditions.