Arc-grounded fault (AGF) in distribution networks significantly threatens personal and equipment safety. In a resonant grounded distribution network (RGDN), the distortion features commonly used for arc detection are altered by frequency characteristics of the equivalent zero mode network, which will pose a massive challenge to ensuring the effectiveness of AGF detection and faulty feeder selection. For these issues, a significant harmonic-based approach for AGF detection and faulty feeder selection in RGDN is proposed in this paper. For more intuitive, compatible, and easily scalable arc signatures, the amplitude-frequency-phase characteristic of principle harmonics in arc current is analyzed in detail with multi-model simulation, and the quantitative relations of phase and amplitude between significant harmonics and industrial frequency (IF) components have been built for constructing the general form of the features set. Then, to be aware of the effect of the Petersen coil on the harmonics in the zero-sequence current (ZSC) of a faulty feeder, the equivalent model of RGDN with its frequency characteristic has been built and analyzed in detail, and the estimation equation of current at fault point is derived. Furthermore, a Taylor- Fourier transform (TFT) based features extraction approach and a targeted strategy of AGF detection and faulty feeder selection have been designed. With simulation data, the effectiveness of the proposed method has been verified.