The invention provides an adaptive water power set fault characteristic extraction method based on multiwavelets. According to the method, peculiar adaptive characteristics of adaptive multiwavelets are utilized, in combination with sensitivity evaluation indexes, optical extraction of water power set fault characteristics is realized. Multiwavelet analysis is better than wavelet analysis in multiple performances, so a multiwavelet analysis method is applied to multiple fields; as multiple scale functions and a wavelet function are comprised, direct acquisition signal decomposition can not be realized during multiwavelet transform, and an acquired signal is required to pre-process and is converted into vector data which can be processed through multiwavelet transform; signal distortion can be constantly caused by an improper pre-processing method to influence extraction of signal characteristics, so selecting a pre-processing method suitable for processing water power set fault signals is crucial for acquiring an accurate and effective multiwavelet decomposition result and highlighting water power set fault characteristics.