State of Health Estimation for Lithium-Ion Batteries Based on Gray Wolf-Optimized CNN-BiLSTM-Attention Hybrid Model

The accurate state of health (SOH) estimation of lithium-ion batteries is essential for enhancing the reliability and safety of battery systems. This paper proposes a battery SOH estimation method based on a hybrid CNN-BiLSTM-Attention model integrated with the Gray Wolf Optimization algorithm. The first step involves selecting capacity degradation-related health factors based on lithium-ion battery charge–discharge curves, followed by the detection of local outliers within this health factor data via the LOF method. Second, the efficacy of each health factor was assessed using Pearson's and Spearman's correlation analyses, followed by normalization of validated factors to construct the HF data set. After that, the Attention algorithm enhances the CNN-BiLSTM architecture, forming a CNN-BiLSTM-Attention hybrid model for adaptive feature weighting. The Gray Wolf Optimizer further tunes its hyperparameters. Finally, the experimental results based on the NASA and CALCE lithium battery data sets demonstrate that the mean squared error, mean absolute percentage error, and root-mean-square error of this method are all within 0.8%, and the coefficient of determination is improved by 0.5 to 1.5%. These results are far lower than those of other methods, demonstrating the prediction accuracy of the proposed model.