Transient diagnosis of induction generators via atom-based time-frequency transforms

Induction generators used in windmills usually work in non-stationary conditions, due to the varying speed of the wind. Therefore, in order to perform its diagnosis through the stator current analysis, adequate time-frequency transforms must be used. Discrete Wavelet Transform succeeds detecting the highest amplitude sideband harmonic caused by asymmetries. Nevertheless, it fails with the rest of faulty components. This paper proposes the use of a continuous time-frequency transform. It avoids the cross-terms introduced by the Wigner Ville Distributions. Moreover, it obtains the right time-frequency resolution needed in each zone of the plane, without the Short Time Fourier Transform or the Continuous Wavelet Transform restrictions. The proposed methodology is defined and experimentally validated: up to four faulty components are detected when diagnosing stator and rotor asymmetries.