High Impedance Grounding Fault Line Selection Based on Fault Phase Residual Voltage Analysis and HBA-XGBoost

Background: As distribution networks expand and become more complex, risks to operational reliability are increasing. Rapid, accurate, and efficient ground fault line selection is crucial for ensuring grid safety, stability, and a reliable power supply. To address the poor identifiability of high impedance grounding faults and the neglect of residual voltage in the fault phase, a high impedance grounding fault line selection method based on fault phase residual voltage analysis and HBA-XGBoost is proposed. It improves fault line identification accuracy and robustness, supporting rapid self-healing and safe operation of complex distribution networks. Methods: This study adopts a three-step methodology: (1) The feedback mechanism of phase residual voltage-current in high impedance grounding faults was analyzed, and interphase imbalance parameter along with 13 auxiliary parameters were extracted to construct a feature database; (2) The Honey Badger Algorithm (HBA) was employed to systematically optimize the hyperparameters of Extreme Gradient Boosting (XGBoost), while conducting feature importance analysis to enhance the overall performance of the model; (3) The optimized HBA-XGBoost fault line selection model was developed to achieve precise grounding fault line identification. Results: To evaluate the adaptability and robustness of the proposed HBA-XGBoost fault line selection model under complex conditions, three tests are conducted: (1) varying signal-to-noise ratios to simulate field disturbances, (2) random data loss to assess performance with incomplete data, and (3) different grounding configurations to emulate topological changes. Results show the model consistently achieves high accuracy and reliable fault identification, confirming its robustness and practical applicability. Discussion: Through a comparative analysis with three existing fault line selection models—Attention- CNN, KPCA-ELM, and SVD-SVM—the proposed method demonstrates notable improvements in accuracy by 10.76%, 5.65%, and 16.18%, respectively. Moreover, in terms of several key performance metrics, including Accuracy, Average accuracy, Average recall, Average F1 Score, and Kappa coefficient, the proposed model consistently outperforms the benchmark models, exhibiting significant overall advantages. Conclusion: This study proposes a fault line selection method based on fault phase residual voltage analysis and the HBA-XGBoost. By integrating mechanism-based modeling with feature optimization, extensive simulation experiments and comparative analyses have substantiated the proposed method’s effectiveness and applicability in practical engineering scenarios. Future research will further investigate collaborative fusion mechanisms for multi-source fault information, combined with adaptive learning architectures, to achieve enhanced fault line selection in distribution networks.