Abstract: A new battery capacity calibration method was proposed to estimate the state of charge (SOC) of power batteries online with combining an unscented Kalman filter (UKF) algorithm. The actual capacities of the batteries were calibrated according to temperature coefficient and current probability, a second-order RC equivalent model was established to simulate the dynamic response characteristics of the battery, and to call the unscented Kalman filter (RLS-UKF) algorithm of recursive least squares with forgetting factor to estimate battery SOC online. The estimation process considered the influence of the change of the OCV (open circuit voltage) -SOC curve caused by the sudden current change. The DST (dynamic stress test) working conditions and the constant current conditions were used to verify the algorithm’s accuracy through MATLAB simulation model. The results show that the RLSUKF algorithm has an average error of 1.2% at DST cycle conditions, and of 1.41% at constant current discharge conditions. Therefore, the method has a better prediction effect than the current mainstream prediction methods.

Key words: power battery, state of charge (SOC), unscented Kalman filter (UKF), recursive least squares (RLS); temperature coefficient, current probability, second-order RC equivalent model