A multi-model probability SOC fusion estimation approach using an improved adaptive unscented Kalman filter technique

Battery model is crucial for the accurate estimation of the state of charge (SOC) in a battery management system of electric vehicles. However, differences exist within optimal battery models corresponding to different types of batteries. Even for the same type of battery, the corresponding optimal battery model may vary with the change of the battery status. To solve the problem, this paper proposes a multi-model probability fusion estimation (MMPFE) method to realize an accurate description of battery characteristics and a precise SOC estimation. An improved adaptive unscented Kalman filter (AUKF) approach is developed for measurement noise variance online update based on the idea of orthogonality between residual and innovation during the SOC estimation. Finally, the proposed MMPFE method was verified by experiments using LiFeO4 and LiMnO2 batteries, respectively. Results indicate that when a voltage drift of +3 mV was applied on the LiFeO4 battery under UDDS condition and an initial SOC error was applied on LiMnO2 battery under FUDS condition at different temperatures, the proposed method still can estimated the precise SOC. Comparing with the results obtained by the other methods under the same conditions, the method presented in the paper shows a higher SOC estimation accuracy and better robustness. •Fusion estimation method based on multi-model considering hysteresis is proposed.•The idea of orthogonality between residual and innovation is employed.•The variance of process noise and measurement noise is updated independently.•The fusion estimated noise variances are fed back to each model channels input.•Dual scale framework is adopted for estimation of parameters and state of charge.