Fusing Temperature-Guided Attention with LFGWO-Optimized CNN-BiLSTM for Accurate Battery SOC Estimation

Precise State of Charge (SOC) estimation for lithium-ion batteries is critical for dependable Battery Management Systems (BMS). However, high precision remains difficult under complex dynamic loads and significant temperature variations. This study proposes a novel deep learning framework integrating a hybrid attention mechanism and an Improved Grey Wolf Optimizer (LFGWO) to address these challenges. The model foundation is a CNN-BiLSTM network enhanced with residual connections. The architecture uses a Convolutional Neural Network (CNN) to extract local temporal features and a Bidirectional Long Short-Term Memory (Bi-LSTM) network to capture long-range data correlations. Addressing temperature effects, we developed a temperature-guided hybrid attention mechanism that dynamically modulates feature weights via channel and temporal attention pathways. Furthermore, a weighted Huber loss function bolsters learning efficacy, while the LFGWO algorithm optimizes hyperparameters. Validated on a comprehensive multi-temperature dataset (e.g., US06, HWFET), our proposed method exhibits superior SOC estimation accuracy, achieving mean MAE and mean RMSE of 0.94% and 1.09%, respectively. The model shows exceptional robustness, particularly in low-temperature environments, and markedly surpasses baseline methods across most tested conditions.