To further improve the accuracy of intelligent diagnosis of mechanical faults of power transformer

the vibration signals of power transformer are symbolized to build the mechanical fault diagnosis model based on stack auto encoder.Firstly

the feature vector of transformer vibration signal was extracted by time series symbolization.Then the complex network is built to extract the feature vector of vibration signals based on degree distribution.The mechanical fault diagnosis model of transformer based on SAE-based is trained and constructed.The analysis results of the vibration signals of a 10 kV dry-type transformer under normal and typical mechanical faults have shown that the dynamical features of vibration signals of transformer can be better described by the symbolization approximation and degree distribution in complex network.The mechanical fault model based on the stacked auto encoder based on SAE has high recognition rate of 95% or above.The results could provide new ideas for the mechanical fault diagnosis of power transformer.