Lithium-ion battery state of charge estimation based on VFFRLS-SVD-AMIUKF

Abstract As the key quantitative indicator of available energy capacity, lithium-ion battery SOC serves as the decision-making basis for optimizing power flow control in marine electrical distribution systems, and within the context of all-electric ship propulsion systems, real-time SOC estimation accuracy directly impacts the effectiveness of power distribution optimization and fault prevention mechanisms. To mitigate the inaccuracies in SOC estimation arising from static battery model parameters and the algorithmic instability of conventional Unscented Kalman Filter (UKF) due to covariance matrix degeneracy, this work introduces a synergistic estimation framework integrating adaptive parameter tuning and a modified UKF variant. This algorithm implements a two-stage estimation architecture: first, the Variable Forgetting Factor Recursive Least Squares (VFFRLS) module performs real-time battery model parameter adaptation; second, an enhanced Unscented Kalman Filter variant-the Adaptive Multi-innovation UKF (AMIUKF)-leverages Singular Value Decomposition (SVD) for robust SOC estimation. MATLAB-based comparative simulations reveal that, when contrasted with conventional lithium-ion battery SOC estimation methodologies, the proposed algorithm demonstrates superior precision in both dynamic model parameter calibration and state-of-charge tracking. This results in enhanced SOC estimation accuracy across varied operational profiles.