Abstract: This study explores a multi-modal data fusion technique for DC circuit defect diagnosis. Taking the DC power supply system renovation project of a 110 kV intelligent substation as an example, a technical solution based on electrical, thermodynamic, and high-frequency oscillation data is designed and implemented. After engineering verification, it was found that this scheme can significantly improve the localization efficiency of DC circuit defects. This technical scheme can accurately locate defects within 30 s after a fault occurs, with a localization efficiency improvement of over 99% compared to traditional methods. Moreover, the recognition rate of high impedance grounding defects can be increased to 92.3%. Therefore, the research on DC circuit defect diagnosis can effectively solve the problems of low efficiency and high missed diagnosis rate of traditional detection methods, and has certain promotion significance for the operation and maintenance management of intelligent substations.