The invention discloses a ferromagnetic resonance fault detection method for an isolated neutral system. Three-phase voltage and a zero-sequence voltage signal of a bus are analyzed to detect the ferromagnetic resonance fault of the system. According to the method, with Hilbert Huang transform (HHT) based on empirical mode decomposition (EMD) as the core, in combination with FFT, FFT analysis is firstly performed on the three-phase voltage and the zero-sequence voltage signal, then HHT analysis is performed, and finally the moment of ferromagnetic resonance and the amplitude value of ferromagnetic resonance over-voltage are obtained through a three-dimensional Hilbert spectrum, so that the ferromagnetic resonance type, the amplitude value and the inducing moment are accurately analyzed in a quantified mode. The HHT signal analysis method is adopted for completely getting rid of linearity and stability constraints, namely, frequency-domain analysis can be performed on the voltage signal, and meanwhile time-domain analysis can be performed; in the process of HHT analysis, an RBF-point symmetry continuation method is adopted, in combination with a mirror image continuation method, the end effect of HHT is improved, and mode superposition is avoided.