During the operation of a transformer, vibration may occur due to factors such as winding stress and magnetic striction of the iron core. The vibration signal of the body can reflect its operating status. Extracting and analyzing the characteristics of vibration signals can assist in achieving transformer health status warning. This article calculates the 50Hz and 100Hz power ratios, low-frequency effective values, high-frequency effective values, and odd and even harmonic power ratios of the main body vibration signals of transformers under different operating states to form feature vectors, and applies support vector machine algorithm to classify the operating states of transformers. And based on this, a monitoring device was developed to achieve onsite detection and classification of transformer operation status, greatly reducing the workload of onsite maintenance personnel and improving their work efficiency.