Robust Model Parameter Identification and SOC Estimation for Li-Ion Batteries Considering Noisy Measurement

The estimation model for state of charge (SOC), which integrates model parameterization with filters, is characterized by elevated precision and robustness. However, unexpected noise perception (outliers) in real-world scenarios may lead to biased model parameterization results, consequently affecting SOC estimation accuracy. This study proposes a novel model-parameterized SOC filtering method that integrates the advantages of kernel width adaptive (KwA) recursive maximum total correntropy (RMTC) and generalized maximum correntropy criterion (GMCC) square root cubature Kalman filter (SRCKF). Specifically, we introduce an efficient RMTC tailored for temporal information processing, incorporating error information induced by input/output noise to enhance model parameterization accuracy. Building on this, the integration of error information into the KwA mechanism maximizes the utilization of error data for adaptive adjustment of kernel width, thereby further improving RMTC’s convergence speed and robustness. In addition, the GMCC containing higher order moments of error distribution is employed as the cost function in SRCKF framework, establishing an effective nonlinear regression model based on a model-driven approach and integrating noise-induced error information into GMCC to enhance SRCKF’s robustness. Finally, the robustness of the proposed method under various types of noise is validated through simulated experimental data.