Many researches have focused on grounding system defects in cross-bonded high voltage（HV） cables in recent years

but the results in previous efforts have not been satisfactory when only using the amplitude of grounding current as indicators. In this paper

the 27 common grounding system states were divided into 3 categories from the perspective of circuit topology

the grounding current calculation method under various defect states was given

and a comprehensive feature extraction based on grounding current amplitude ratio and phase angle difference method was proposed. The linear binary classification logistic regression algorithm was introduced into the intelligent identification model of cross-connected high-voltage cable grounding system defects through polynomial feature expansion and multi-class voter construction

and the classification algorithm was evaluated using the data generated by the grounding current calculation model. The evaluation results show that the grounding system defect classification method based on the circuit topology can be adopted to effectively distinguish various types of defects

and the comprehensive grounding current characterization method based on the amplitude ratio and the phase angle difference can be adopted to effectively reflect the abnormal circuit topology of the grounding system. The system grounding defect classification model based on logistic regression can be adopted to effectively identify 27 common high-voltage cable grounding system states. The total evaluation indicators such as algorithm accuracy

precision

recall and F1 value are all above 95%. The recognition accuracy of the model for various states is above 89%. The research results verify the effectiveness of the proposed method.