In order to further study the large amount of fault information contained in transformer vibration signals, a fault diagnosis method for transformer winding looseness based on S-transform singular value decomposition (ST-SVD) and support vector machine optimized by whale optimization algorithm (WOA-SVM) model was proposed. Firstly, based on the transformer fault simulation experiment platform, the vibration signals of transformer windings in different states were collected. Secondly, the time-frequency matrix of the transformer vibration signal was obtained by S-transformation. Thirdly, calculating the amplitude matrix corresponding to the time-frequency matrix for SVD, and defining the feature vector. Finally, the parameters of SVM model were optimized by whale optimization algorithm, and the fault diagnosis was completed by inputting feature vectors. The experimental results show that the accuracy of fault identification of the proposed method is higher than that of the traditional method model, and it is suitable for the diagnosis of transformer winding looseness fault.