The invention relates to the field of wind power generation equipment, in particular to a fault detection method for a wind turbine generator unit. The wind turbine generator unit applied to the method comprises a base, a stand column, a diagonal pole, a power generation mechanism, a switch, a driving unit, a telescoping unit, a retaining ring, a first air bag, a stop block and a controller. The method has the advantages that sounds of the wind power generator unit during operation are acquired by a sound detecting device, and are transformed into a marginal spectrum through Fourier transformand Hilbert Huang transform, and the marginal spectrum is compared with a standard reference curve by a computer to determine whether the wind turbine generator unit is abnormal or not, so that maintenance personnel is prevented from frequent detection in desert areas, and the timeliness of detection of the wind turbine generator unit is improved.