Abstract: In view of the complex transient characteristics of short-circuit faults in DC power network and the

difficulty of fault identification，the short-circuit faults in DC power network can be identified by detecting the change of capacitance voltage through the analysis of the VSC topology and switch characteristics. A short-circuit fault identification method based on DC-side capacitive voltage wavelet approximate entropy was presented，which is used to train BP neural network. The simulation results show that the characteristics of time-frequency localization and approximate entropy of wavelet variation are combined with the characteristics of approximate entropy of wavelet to describe the transient signal. The fault features can be extracted accurately and the short-circuit faults of DC power network can be identified accurately and quickly.

      Key words: DC power network；fault identification；wavelet transform；approximate entropy；BP neural network