A Novel Approach to Micro-Short-Circuit Fault Detection in Lithium-Ion Batteries via NGO-Optimized VMD and XGBoost Integration

With the widespread use of lithium-ion batteries, early fault diagnosis is crucial for the safety and lifespan of the batteries. To address the issue of micro-short-circuit faults being difficult to detect, this paper proposes a fault diagnosis method based on NGO-optimized VMD and XGBoost integration. This method optimizes the mode decomposition number (K) and secondary penalty factor (α) of VMD using NGO, extracts features from the battery voltage signals, and inputs them into the XGBoost model for fault diagnosis. The simulation results show that the VMD-XGBoost model achieves approximately 3% higher accuracy compared to the EMD-XGBoost model, while the NGO-optimized VMD-XGBoost model has an average diagnostic accuracy exceeding 90% across all scenarios, significantly outperforming traditional methods. Further comparative experiments show that, under the NGO-optimized VMD, XGBoost outperforms other models such as DT, RF, and SVM in micro-short-circuit fault diagnosis. Further comparative experiments show that, with NGO-optimized VMD, XGBoost outperforms other models such as DT, RF, and SVM in micro-short-circuit fault diagnosis, validating the effectiveness and reliability of the proposed method, which also holds promising application prospects in the field of micro-internal short-circuit fault diagnosis for lithium-ion batteries.