Abstract: Accurate assessment of the state of health (SOH) for lithium-ion batteries is critical for ensuring the safe and stable operation of equipment such as emergency power supplies. To address the limitations of traditional methods, such as inaccurate feature extraction and heavy reliance on extensive test data, this study proposes a SOH estimation method integrating multi-feature analysis with an LSTM-XGBoost hybrid model. The long short-term memory (LSTM) network is utilized to predict feature data for future cycles, mitigating data dependency; the eXtreme Gradient Boosting (XGBoost) algorithm establishes the mapping relationship between features and SOH for efficient estimation. Experiments conducted on the NASA battery aging dataset demonstrate that the proposed method achieves a root mean square error (RMSE) below 1% and a mean absolute error (MAE) less than 0.8% in SOH estimation, significantly outperforming benchmark models, indicating high accuracy and strong robustness.