| Title |
High-Accuracy SOC Estimation of LFP Batteries via Optimal OCV Derivation Using a Discrete Preisach Model and Extended Kalman Filter |
| Authors |
안종찬(Jong-Chan An) ; 송민우(Min-Woo Song) ; M.J.M.A Rasul(M.J.M.A Rasul) ; 김종훈(Jong-Hoon Kim) |
| DOI |
https://doi.org/10.5370/KIEE.2026.75.5.1068 |
| Keywords |
Lithium iron phosphate; Discrete Preisach model; Hysteresis; State of charge; Battery management system |
| Abstract |
Lithium iron phosphate (LFP) batteries are widely used in electric vehicles and energy storage systems because of their thermal safety and low cost. However, their intrinsic voltage hysteresis complicates accurate state-of-charge (SOC) estimation by creating different voltage responses during charge and discharge. This study proposes an SOC estimation method that combines a discrete Preisach model (DPM) with an equivalent-circuit-model-based extended Kalman filter(EKF). The proposed method identifies DPM parameters using weighted least squares and ridge regularization with a bias term, improving numerical stability and reducing hysteresis-induced uncertainty under dynamic operating conditions. To evaluate practical applicability, dynamic stress test (DST) and federal urban driving schedule (FUDS) profiles were applied at 25 °C, 45 °C, and 0 °C. Experimental results showed that the proposed EKF+DPM(WLS+Ridge+bias) achieved the lowest mean absolute error (MAE) at all temperatures, confirming improved SOC estimation accuracy in terms of average error. In particular, at 0 °C, the MAE decreased from 3.03% to 1.45% under DST and from 3.61% to 0.58% under FUDS compared with the conventional EKF. By contrast, the conventional EKF+DPM approach increased the maximum error and MAE under some conditions, especially at low temperature, indicating that stable DPM parameter estimation is critical to performance. These results demonstrate that the proposed method can stably suppress SOC estimation errors despite temperature variation and dynamic load changes, providing improved reliability and robustness for real-time battery management system applications. |