There is a technical problem that with a large amount of harmonic monitoring data in the power grid it is difficult to realize the resonance monitoring of the power grid because of the lack of phase information. A set of grid resonance source identification methods considering dominant coefficients, correlation coefficients and sensitivity indices is proposed. First, based on a multi-harmonic source resonance analysis model, the correlation and regression characteristics of the harmonic parameters in the resonance state are analyzed, and a strong dominant resonance state identification method taking into account the dominant coefficients is proposed. Secondly, the harmonic impedance and data characteristics in the strong dominant resonance state are demonstrated, and the resonance state of the power grid and the resonance source identification method integrating the mechanism model and the resonance data characteristics are constructed. Finally, based on modeling simulation and engineering cases, it is verified that the coupled harmonic impedance measurement deviation of the sensitive resonant branch and the sensitive harmonic source is less than 2% in the condition that the background harmonic contribution to the harmonic voltage of the bus is less than 20%. The proposed method is able to accurately identify the resonance state and the resonance source.