State-of-Charge Estimation of Lithium-Ion Batteries Under Low-Temperature Conditions

Low-temperature operation intensifies polarization, causes model-parameter mismatch, and changes the time scale of the dynamic response in lithium-ion batteries. These effects reduce the accuracy of state-of-charge (SOC) estimation. To address this problem, an offline estimation method based on improved forgetting factor recursive least squares and three-dimensional (3D) bridge compensation is proposed. A 25 °C open-circuit voltage (OCV)-SOC curve is used as a unified reference. Under a first-order resistor–capacitor (RC) model, SOC segmentation and a variable forgetting factor are introduced for parameter identification. A genetic algorithm and an extended Kalman filter are then combined to construct the 3D bridge compensation. The results show that the proposed method effectively reduces SOC estimation errors and improves terminal-voltage reconstruction. Under the calibration conditions, the average root mean square error (RMSE) and mean absolute error (MAE) of SOC estimation, evaluated against the Coulomb-counting reference SOC trajectories, decreased from 0.0895 and 0.0834 to 0.0031 and 0.0021, respectively. The RMSE and MAE of terminal-voltage reconstruction, evaluated against the measured terminal voltage, decreased from 0.1623 V and 0.1113 V to 0.0325 V and 0.022 V, respectively. The proposed method provides an improved solution for offline SOC estimation of lithium-ion batteries at low temperatures. It also shows a degree of cross-cycle applicability for the same cell within the calibrated temperature range and provides a reference for correcting the state-estimation accuracy of fixed-reference models.