In order to diagnose transformer winding looseness fault more accurately and effectively, a fault diagnosis method for transformer winding looseness based on variational mode decomposition (VMD) and support vector machine optimized by whale optimization algorithm (WOA-SVM) is proposed. Firstly, a fault simulation experiment is carried out on a 10 kV transformer to measure its vibration signal. Then, VMD is used to decompose the non-stationary vibration signal into multiple intrinsic mode functions (IMF), and the energy entropy of each IMF is calculated to constitute feature vectors. Finally, the feature vectors are input into the WOA-SVM to train the classification model, and the fault diagnosis of transformer winding looseness is realized. The results show that the proposed method is applicable to the fault diagnosis of transformer winding looseness, and its fault identification accuracy is higher than the traditional improved SVM classification model.