State‐of‐charge estimators considering temperature effect, hysteresis potential, and thermal evolution for LiFePO4 batteries

• Published in: International journal of energy research Vol. 42; no. 8; pp. 2710 - 2727
• Main Authors: Xie, Jiale, Ma, Jiachen, Bai, Kun
• Format: Journal Article
• Language: English
• Published: Bognor Regis John Wiley & Sons, Inc 25.06.2018
• Subjects: adaptive extended Kalman filter; Adaptive filters; Algorithms; Batteries; Computer simulation; differential evolution algorithm; Electric potential; Equivalent circuits; Estimators; Evolutionary algorithms; Exploitation; Extended Kalman filter; Heat generation; Hysteresis; hysteresis model; Hysteresis models; Internet; Kalman filters; Lithium-ion batteries; Mathematical models; Parameter estimation; Parameter identification; Parameters; State of charge; Temperature effects; Thermal evolution; thermal evolution model; Tracking; Voltage
• ISSN: 0363-907X; 1099-114X
• DOI: 10.1002/er.4060

Summary To achieve accurate state‐of‐charge (SoC) estimation for LiFePO4 batteries, the effects of temperature, hysteresis, and thermal evolution are elaborately modeled. Open‐circuit voltage is regarded as the sum of electromotive force and hysteresis potential (Vh), where electromotive force is constructed as the function of SoC and temperature and Vh is reproduced with a geometrical model. By simulating battery heat generation and dissipation, a thermal evolution model is established and exploited for open‐circuit voltage and parameter identification. Then, on the basis of a second‐order equivalent circuit model, 2 SoC estimation schemes are proposed: One scheme uses the recursive least square with forgetting factor algorithm and off‐line equivalent circuit model parameters derived by the differential evolution algorithm; the other scheme resorts to the adaptive extended Kalman filter (EKF) and online tuned parameters. Experiments validate the effectiveness of the hysteresis model and the thermal evolution model. In contrast to a joint EKF estimator, experimental results under different temperatures and initial states suggest that both the proposed estimators are superior to the joint EKF estimator. Benefiting from the online updated parameters, the adaptive EKF estimator behaves best for giving consistent SoC‐tracking performance under different conditions. A battery thermal evolution model is formulated and exploited for OCV prediction and parameter identification. An off‐line approach using the differential evolution algorithm and an on‐line approach based on an LMS filter are proposed for ECM parameterization. The RLSF and AEKF algorithms are used to estimate the SOC.