Aiming at the problem of transformer DC bias caused by the urban rail transit, this paper proposes a DC bias state identification method based on vibration signal for transformers of urban power grid. Firstly, the time of duration and the frequency characteristics of the transformer vibration signal under the conditions of DC bias, short fault and harmonic interference is analyzed. It is found that compared with other faults, the transformer vibration intensifies under DC magnetic bias. The frequency component of vibration signal becomes complex. A series of high-order harmonic components, especially the 50Hz odd double frequency component, increase significantly. Based on these phenomena, the influence of short circuit fault on DC bias state identification is eliminated by using the sum of energy of 50 Hz octave component of vibration signal except 100 Hz. Then, the ratio of the sum of the energy of 50 Hz odd octave components except 100 Hz over the sum of energy of 50 Hz octave components except 100 Hz is used to eliminate the influence of power grid harmonic interference on DC magnetic bias state identification, so as to realize the state identification of transformer DC magnetic bias caused by urban rail stray current. Finally, the on-site measured data is analyzed and processed to further verify the accuracy of proposed method.