A Novel Adaptive Particle Filtering–Gated Recurrent Unit Correction Framework for Multi-State Cooperative Estimation of Lithium-Ion Batteries

Abstract

The accuracy of state-of-charge (SOC), state-of-energy (SOE), and state-of-health (SOH) estimations directly affects the system’s ability to sense the battery's state. Therefore, this paper proposes a multi-state co-estimation method that integrates adaptive particle filtering with gated recurrent unit (GRU) error correction. First, a second-order equivalent circuit model considering temperature and hysteresis characteristics is constructed. Noise modeling and an adaptive mechanism are employed to optimize the recursive least squares method, improving the adaptability of parameter identification. Second, to address the sample degradation and depletion issues of particle filtering, an adaptive particle filtering method integrating Kalman filtering and optimization strategies is constructed, improving the accuracy and convergence performance of SOC and SOE estimation. Furthermore, considering the differences in multi-state time-varying characteristics, a dual-timescale co-estimation framework incorporating GRU dynamic error correction is established to enable information exchange and closed-loop updates for SOC, SOE and SOH. Experimental validation at different temperatures and ageing stages indicates that the root-mean-square-error (RMSE) for parameter identification ranges from 8.26 to 26.86 mV; in the case of co-estimation, the RMSE for SOC and SOE ranges from 0.232% to 0.547%, whilst that for SOH ranges from 0.001% to 0.008%. This indicates that the method has a relatively stable estimation capability.