SOC Estimation of Lithium Batteries Based on adaptive sliding window filtering and PKO-HKELM

The state of charge (SOC) estimation of lithiumion batteries is critical for battery management systems (BMS), as it directly impacts electric vehicle drivingrange prediction and energymanagement efficiency. To overcome challenges such as sensornoise interference and inadequate dynamic modeling under transient conditions, this paper proposes an optimized framework based on a hybrid kernel extreme learning machine (HKELM) with integrated dynamic noise filtering. First, an adaptive slidingwindow filtering (ASWF) method is employed to suppress highfrequency noise in voltage and current measurements. Next, a pied kingfisher optimizer (PKO) is developed to finetune the parameters of the hybrid kernel, thereby enhancing the model’s nonlinear fitting capability. Finally, the ASWF-PKO-HKELM model was verified using the A123 APR18650M1A lithium iron phosphate battery dataset under the Federal Urban Driving Schedule (FUDS), Dynamic Stress Test (DST), and Urban High Speed (US06) driving conditions. The experimental results show that the method has an average absolute error (MAE) of only 0.102% and a root mean square error (RMSE) of 0.117%, outperforming existing methods in terms of accuracy and robustness.