Dual‐Layer Ensemble Learning Framework for Lithium‐Ion Battery State of Health Estimation

Lithium‐ion batteries, as critical energy storage components in electric vehicles and energy storage systems, require accurate state of health (SOH) estimation to ensure operational safety and extend service life. However, traditional data‐driven methods often suffer from insufficient robustness when facing complex degradation behaviors, small‐sample scenarios, and noisy environments. To address these challenges, this article proposes a dual‐layer ensemble learning framework based on bidirectional long short‐term memory–modified Newton–Raphson‐based optimizer–convolutional neural network with dual‐attention mechanism (BiLSTM [long short‐term memory]–MNRBO–CNN–Dual Attention) for lithium‐ion battery SOH prediction. First, multidimensional health factors (HFs) are extracted from four complementary characteristic curves derived from battery charge–discharge data. The Pearson correlation coefficient (PCC) and Spearman correlation coefficient (SCC) analyses are then employed to select degradation‐sensitive features. Subsequently, multiple BiLSTM sub‐models are constructed to capture temporal degradation patterns. An MNRBO is further introduced to adaptively determine optimal fusion weights. Finally, a CNN–Dual Attention meta‐learning module achieves nonlinear feature mapping and improves prediction accuracy and model robustness. Experimental results on the Oxford, National Aeronautics and Space Administration (NASA), and laboratory aging datasets demonstrate that the proposed method achieves superior performance compared with conventional CNN, LSTM, and single‐layer ensemble models. The proposed framework maintains both mean absolute error (MAE) and mean absolute percentage error (MAPE) within 1%, demonstrating high prediction accuracy and strong generalization capability.