A characteristic energy method of the power distribution network grounding real-time fault detection is characterized by selecting a Daubechies8 wavelet filter to carry out the 4-order wavelet decomposition on a power distribution network fault recording zero sequence current time series; taking the integrated square of the fourth-layer wavelet decomposition coefficients as the characteristic energy of the fault monitoring, and determining whether a power distribution network grounding fault occurs according to the energy characteristic value of a zero sequence current; when the grounding fault occurs, searching a monitoring point having the largest characteristic energy in the feeder line branches as the fault starting point, and along with the flow-out bus directions of the power flow of the branches, searching a monitoring point having the smallest adjacent characteristic energy as the fault tail, thereby determining that a power distribution network grounding fault section is between the above head and tail end points.