A dual-exponential EKF model for lithium-ion battery RUL prediction with Bayesian optimization

Abstract Accurate prediction of the remaining useful life (RUL) of lithium-ion batteries is essential for the safe and efficient operation of electric vehicles and energy storage systems. This study proposes a novel RUL prediction framework that combines a physically interpretable dual-exponential degradation model with an Extended Kalman Filter (EKF), enhanced through Bayesian optimization. The proposed method automatically tunes critical EKF parameters ( P , Q , and R ), addressing a key limitation of conventional EKF approaches that rely on expert-driven parameter settings. Additionally, a dynamic weighting loss function is introduced to improve robustness across different degradation stages, emphasizing capacity trend consistency in early life and RUL accuracy near failure. Validation is performed on two widely used datasets (CALCE and NASA) across four experimental configurations. Results show that the proposed model outperforms standard EKF, Seq2Seq, and other baseline models in both capacity tracking and RUL estimation, demonstrating superior accuracy, interpretability, and generalization across diverse battery types and aging patterns. This work provides a scalable, interpretable, and measurement-driven approach for battery health management in real-world applications.