A Novel Particle Filter with Rank-Sampling Based on Adaptive Parameter Identification for Lithium-Ion Battery SOC Estimation

Particle degeneracy remains a critical challenge in conventional particle filter (PF) algorithms for battery state-of-charge (SOC) estimation, leading to reduced estimation accuracy and filter divergence. To address this limitation, this study proposes an adaptive rank particle filtering (ARPF) algorithm that integrates a rank-sampling mechanism to enhance particle diversity and estimation robustness. First, an improved adaptive forgetting factor recursive least squares (IAFFRLS) method is developed for parameter identification of a second-order RC equivalent circuit model. The forgetting factor is dynamically adjusted based on prediction error, thereby improving both identification accuracy and adaptability to time-varying battery characteristics. Second, a rank particle filter (RPF) is constructed by reformulating the sampling step of the conventional PF through rank-sampling. This approach generates a more reasonable proposal distribution for each particle, effectively mitigating particle degeneracy while preserving particle diversity. Third, building upon the RPF framework, the ARPF algorithm incorporates Sage–Husa adaptive theory to enable real-time adjustment of process and measurement noise covariances, further enhancing the filter’s robustness against model uncertainties and environmental variations. The proposed ARPF algorithm is validated through extensive experiments conducted under multiple operating conditions, including three temperature levels (0 °C, 25 °C, and 45 °C) and two dynamic drive cycles (DST and FUDS). Experimental results demonstrate that the ARPF achieves superior SOC estimation performance, with mean absolute error (MAE) ranging from 0.06% to 0.24% and root-mean-square error (RMSE) ranging from 0.24% to 0.40% across all tested conditions. Comparative analysis reveals that the ARPF outperforms both conventional PF and rank filtering (RF) methods in terms of estimation accuracy, convergence speed, and robustness, particularly under challenging operating scenarios.