An Online Parameter Identification Method for Lithium Batteries Based on Improved Multi-Adaptive Forgetting Factor Recursive Least Squares

Accurate parameter identification of lithium battery equivalent circuit models under dynamic operating conditions remains a critical challenge. To address this, we propose an improved multi-adaptive forgetting factor recursive least squares (IMAFFRLS) algorithm. First, offline parameter identification is conducted via constant-current pulse tests, systematically comparing the performance of a first-order RC model with variable parameters and a dual-polarization (DP) model. Experimental results demonstrate that the DP model achieves superior accuracy in terminal voltage fitting. For online parameter identification, the conventional Forgetting Factor Recursive Least Squares (FFRLS) method is evaluated, revealing its limitations under non-stationary conditions. To overcome these issues, the IMAFFRLS algorithm is introduced, incorporating adaptive forgetting factors and multi-parameter optimization. Validation under dynamic stress test (DST), federal urban driving schedule (FUDS), and US06 driving cycle conditions confirms that IMAFFRLS significantly outperforms FFRLS in terminal voltage prediction accuracy and enhances parameter adaptability across diverse operating scenarios. This work provides a robust solution for real-time battery management systems.