Multi-Factor SOH Estimation for Lithium-Ion Batteries via Attention-GNN and SVR Fusion

To enhance the accuracy of State of Health (SOH) estimation for lithium-ion batteries under complex aging scenarios, a novel model integrating multiple physical factors is proposed. Using the NASA battery aging dataset, this work investigates the degradation behavior and estimation performance under multifactor coupling conditions. The model incorporates a channel attention mechanism and graph neural network (GNN) to extract degradation features and capture non-Euclidean spatial dependencies among variables. Furthermore, Support Vector Regression (SVR) is employed to enable accurate mapping of high-dimensional features to SOH estimates. This design allows the model to effectively integrate heterogeneous signals, including voltage, current, temperature, and strain. Experimental results demonstrate that the proposed approach surpasses traditional methods in terms of single-cell modeling, cross-cell generalization, and full-cycle estimation. Notably, the model maintains superior accuracy and robustness under high-noise conditions and distribution shifts, indicating strong generalization capabilities. These results suggest that the proposed framework offers a promising solution for reliable battery health estimation in real-world applications.