An arcing high-impedance model is presented. An intelligent new approach for arcing high-impedance-grounded fault detection in resonant grounded system is presented conbined Hilbert-Huang transform (HHT) with probabilistic neural network (PNN). Zero-sequence current (including the faulted cycle, the 1/4 before fault and the 3/4 after fault) of each circuit caused by single-phase grounded fault is collected, and then the wave is analyzed by EMD and the characteristic of the intrinsic modal energy is calculated. Then the intrinsic modal energy is put into the PNN network. In this way, fault line selection is achieved automatically. Arcing high-impedance-grounded fault and hybrid cable system grounded fault are simulated by MATLAB in this method and noise interference experiment proves the method is feasible and accurate.