The invention relates to a method for identifying a frequency-converting speed-adjusting dynamic running mode of an electric motor based on time-frequency analysis. The method is characterized by comprising the following steps: generally demodulating an original current signal into non-overlapped components, by comprehensively utilizing a general demodulating method, a MODWPT (Maximal Overlap Discrete Wavelet Packet Transform) decomposing method and a general inverse demodulating method; separating the components according to the MODWPT decomposing method; separating a general demodulating phase function corresponding to a main component of the current; and identifying the running mode of the electric motor according to a curve and an analysis express form corresponding to a differential coefficient of the function. According to the method, the components can be separated from a multi-component time-frequency changing signal, the practical running condition of the electric motor is fully considered in a signal analysis scope, and the frequency analysis express form is automatically finished by a program in the form of curve fitting. Thus, the invention provides the method for accurately identifying the running mode of the electric motor under the dynamic state of the frequency-converting speed-adjusting electric motor based on the time-frequency analysis.