Online estimation of Lithium-ion battery equivalent circuit model parameters using decoupled recursive least squares with adaptive moving window forgetting factor

This paper presents a novel adaptive moving window forgetting factor method specifically designed for the decoupled recursive least squares approach applied to the second-order equivalent circuit model of batteries, The proposed technique adjusts weights using historical and current modeling errors, reducing complexity and improving parameter accuracy. To further improve its efficiency, the jellyfish search algorithm was employed to fine-tune the hyperparameters. Three distinct load profiles—the federal urban driving schedule, the dynamic stress test, and the supplemental federal test procedure driving schedule for United States —were selected to validate the effectiveness of the proposed algorithms. Experimental results show that the proposed method reduces the mean absolute error of terminal voltage by up to 74.3%, with an average improvement of 34.6% over the fixed forgetting factor method. Furthermore, it outperforms the sliding window difference forgetting factor method by up to 52.6%, with an average gain of 27.8%. Lastly, the method demonstrates up to 42.6% improvement and an average enhancement of 26.0% 20 over the original adaptive moving window forgetting factor method. In comparison to the aforementioned three methods, the proposed approach shows the capability to reduce mean absolute error of state of charge by up to 40.5%, 23.0%, and 7.6%, respectively