Abstract: On-load tap changer (OLTC) is one of the most fragile parts in power transformer. In order to realize the fault diagnosis of OLTC, a fault diagnosis strategy based on the barycenter average dynamic-time warping algorithm is proposed in this paper. In accordance with the nonlinearity and strong time-varying of OLTC vibration signal, an adaptive variational mode decomposition algorithm is proposed to calculate the time-frequency characteristic matrix of OLTC vibration signal. Furthermore, the dynamic-time warping barycenter average algorithm is introduced to adaptively calculate the difference between the time-frequency characteristic matrix of OLTC vibration signals, so as to realize the fault diagnosis of OLTC. The results show that the dynamic-time warping barycenter average algorithm can improve the robustness of fault diagnosis by calculating the barycenter average sequence of multiple groups of signals. At the same time, based on the calculated difference, not only the diagnosis of OLTC faults can be realized, but also the severity of the fault can be quantitatively characterized. The research of this paper can provide new ideas for OLTC fault diagnosis.