Abstract: The composition of the double-fed induction generator（DFIG）is complex，and its operation and

maintenance are difficult. Due to the relatively complex electromagnetic transient characteristics of the fault，the

extraction quality directly affects the reliability of the fault detection results of the DFIG. Aiming at the limitations

of single fault extraction，a fault electromagnetic transient response feature extraction method for DFIG was

studied. The transient wave recorder was used to collect the fault electromagnetic transient response signal of the

DFIG and implemented noise elimination and amplification processing. The transient response signal was

decomposed by the empirical mode decomposition（EMD）method，and several image formation model（IFM）

components were obtained，from which energy entropy，kurtosis and center frequency were extracted and

normalized. The results show that the energy entropy measured by the electromagnetic transient response signal

samples of the fault-free DFIG is within 0.9，the kurtosis is within 0.1，and the center frequency is between 0.012

and 0.016，which is smaller than the eigenvalues of the other four groups of samples，indicating that there are

obvious differences in the characteristics of fault signals and normal signals，which provides a reliable basis for

this research.

      Key words: double-fed induction generator（DFIG）；fault electromagnetic；transient response signal；empirical mode decomposition（EMD）method；feature extraction