Abstract: Mining cables are affected by the harsh environment of coal mines, and are prone to insulation degradation and sheath damage. The traditional offline diagnostic methods such as low-voltage pulse method and partial discharge method are often used for detecting mining cables. The methods are complex to operate and have low accuracy, making it difficult to meet the needs of modern coal mine production. However, the existing harmonic based cable fault diagnosis methods have problems such as bulky detection devices, low detection precision, and difficulty in application in coal mines. In order to solve the above problems, a degradation monitoring and diagnosing method of mining cables based on current harmonic features is proposed. The method extracts high-order harmonic content information in cables as fault feature vectors, normalize the feature vectors, and then import them into extreme gradient boost tree (XGBoost) model. Combined with known cable fault degradation value data, a training sample set is formed to train the XGBoost model. Finally, the method uses the constructed XGBoost model to monitor and diagnose cable degradation in real-time. The simulation results show that the relative energy of the extracted high-order harmonic vectors from different parts of the cable is significantly different. The extracted high-order harmonic vectors can characterize the operating status of different parts of the cable. The goodness of fit parameter R² of the XGBoost model is as high as 0.93, and the error is small. The case analysis results verify that the degradation monitoring and fault diagnosis method of mining cables based on current harmonic features can provide real-time and accurate monitoring and diagnosis of the operation status and degradation faults of mining cables.