Estimation of State of Health and Knee Point Identification in Lithium-ion Batteries Using BES-GRU

Abstract

Accurately assessing the State of Health (SOH) of batteries and conducting knee point detection is of utmost importance in prolonging their lifespan. However, the generalization ability and robustness of individual methods are limited. To address this, a novel lithium-ion battery SOH estimation model is proposed, incorporating the Bald Eagle Search (BES) optimization algorithm in conjunction with the Gate Recurrent Unit (GRU) neural network. The selection of highly correlated health indicators (HI) with battery SOH is performed using the Pearson correlation coefficient. The BES algorithm is employed to optimize the selection of parameters for GRU networks. The model reliability is verified using the MIT-Stanford lithium-ion battery dataset. The results illustrate that the proposed BES-GRU model effectively predicts the SOH of batteries and identifies knee points, effectively improving the estimation accuracy.