Electrified railway is one of the major harmonic sources in the power system, which is greatly affected by the route and weather. So, precise detection of electrified railway power load harmonic is of great important for the safety of power grid. A math mode of electric locomotive was built in this paper, from which the load current of the railway under different conditions was acquired. An improved HilbertHuang transform was applied to extract the harmonic of the current waveform in which bandwidth restricted empirical mode decomposition (EMD) method was used. By adding masking signal, the mode mixing could be restrained effectively. The simulation results show that the proposed method can extract the time-frequency distribution of the harmonics in electrified railway, which provides a new method for harmonic detection of electrified railway.