Abstract:Transformer vibration analysis is still one of the hot spots in the fault diagnosis of transformers in operation.In order to study the influencing factors of transformer vibration,This article collects the vibration signals of some transformers under different operating conditions, Through big data fit training and frequency domain diagram feature analysis, etc., The effects of different current harmonic distortion, power factor, temperature, and operating life have been studied, The correction curve of the influence of current distortion rate on the spectrum of vibration signal and the repair formula of temperature on vibration amplitude are obtained.On the basis of the correction curve, correction formula and analysis of a large amount of data proposed in this article, Aiming at the problem of transformer core loose fault, a transformer core loose fault diagnosis method based on vibration signal is proposed, that is, a feature quantity extraction model is established by the changes of frequency components of 50Hz, 150Hz, 100Hz and 200Hz.The transformer state is judged by comparing the characteristic vectors under normal and loose conditions,Finally, the effectiveness of the diagnostic method is verified by analog test.