Abstract：Three-level neutral-point-clamped（3L-NPC）interconnected converters have been widely used in

the AC-DC hybrid distribution grids due to their superiorities of large capacity and high power quality. However，

their working conditions are always with high power，varying load，and limited heat dissipation，etc.，with a high

open-circuit failure rate of power switches. Meanwhile，existing fault diagnosis methods are mostly single

mechanism-based or data-based，unable to overcome the problems of complex system model structure and changing operating conditions，resulting in low diagnostic accuracy and speed. To this end，a mechanism-data-fusion-driven fault diagnosis method for interconnected conversion systems was proposed. Firstly，a mechanism-data-fusion model was constructed using a neural network observer to improve the fault diagnosis accuracy. Subsequently，the trajectories of current residuals after open-circuit faults of different devices were analyzed，and a current residual table was summarized，based on which a fast and accurate open-circuit fault diagnosis method was formed. Finally，the experimental and hardware-in-the-loop results verify the effectiveness of the proposed method.

      Key words：AC-DC hybrid distribution grid； three-level interconnected converter； fault diagnosis；

mechanism-data-fusion；current residual