A Bias Correction Based State-of-Charge Estimation Method for Multi-Cell Battery Pack Under Different Working Conditions

In order to estimate the state-of-charge (SoC) for all cells in the battery pack, this paper proposed an average cell model to represent every cell in the pack. The average cell model consisted of a basic model and a bias function. First, the parameter identification of the basic model was conducted, and the inconsistencies between cells were calibrated by the uncertainties of the basic model parameters. Second, artificial neural networks were used to construct the response surface approximate model of the bias function. In order to make the average cell model more adaptable to different working conditions, a novel bias function considering the polarization voltage and the temperature was proposed to correct the basic model, and it was compared with other bias functions. Then, the extended Kalman filtering algorithm was used for SoC estimation based on the corrected model. Finally, a case study with six lithium-ion battery cells was performed for the verification and evaluation of the proposed method. The results indicated that the average model corrected by the proposed bias function showed good adaptability to different working conditions, and the maximum absolute SoC estimate errors of all cells in the battery pack were less than 2% at 25 degrees C, and 3.5% at 10 degrees C or 40 degrees C.