State of charge evaluation of battery in electric vehicles based on data-driven and model fusion approach

To address the problem of large errors in SOC estimation using model methods, this paper integrates data-driven and model-based methods and proposes an online SOC evaluation method based on fuzzy PID optimized unscented Kalman filter (FPID-UKF). This method establishes a fusion method SOC evaluation model architecture, analyzes the data-model fusion evaluation method, and studies the online evaluation method of FPID-UKF. The evaluation method firstly adopts the forgetting factor recursive least square (FFRLS) method to identify the parameters of the equivalent circuit and then, takes the unscented Kalman filter (UKF) to estimate the SOC of the established battery state evaluation model; meanwhile, to further improve the prediction effect of UKF, Kalman gain and innovation were controlled by PID, and the PID coefficient was adjusted by fuzzy control to realize the optimization of UKF by FPID. Finally, the battery data from the University of Maryland were verified under US06 Highway Driving Schedule (US06) and Federal Urban Driving Schedule (FUDS) conditions, and the results indicate that the SOC estimation error does not exceed 3.07% under US06 condition and 4.79% under FUDS condition, and the FPID-UKF algorithm has better estimation accuracy than EKF and UKF.

Automated summary

In order to overcome the large errors in SOC estimation using model methods, this paper proposes an online SOC evaluation method based on fuzzy PID optimized unscented Kalman filter (FPID-UKF) by integrating data-driven and model-based methods. The method establishes a fusion method SOC evaluation model architecture, analyzes the data-model fusion evaluation method, and studies the online evaluation method of FPID-UKF.