Abstract: Based on the importance of state of charge(SOC) to the battery management systems(BMS), a Grey Wolf Optimizer enhanced SVD-based Multi-Innovation Unscented Kalman Filter(GWO-SVD-MIUKF) for SOC estimation is proposed. The method combines Grey Wolf Optimizer(GWO) with a singular value decomposition-based MIUKF(SVD-MIUKF). Both the parameter identification and the filter structure are optimized. SVD is used to reconstruct the covariance matrix in MIUKF to improve nu?merical stability, while GWO is used to identify model parameters and dynamically adjust the es?timation window, enhancing adaptability and convergence. Experiments are conducted on the public INR18650-20R dataset from the University of Maryland under various typical conditions. Results show that the proposed method achieves high estimation accuracy, with SOC error controlled within ap?proximately 0.20%, and demonstrates good convergence performance.

Key words: lithium-ion battery, state of charge(SOC) estimation, improved multi-innovation unscented Kalman filter(MIUKF), GWO-SVD-MIUKF