Skip to Main Content
Online monitoring of the mechanical performance of onload tap changers (OLTCs) within high-voltage (HV) power transformers is of utmost significance for a safe, stable, and reliable operation of the power systems. This paper investigated a novel strategy based on a Hidden Markov Model (HMM) for mechanical fault diagnosis of OLTCs. With partition, normalization, and vector quantization of the power spectral density of the obtained vibration signals, a feature vector extraction methodology was presented for the discrete power spectrums which, to the farthest extent, could retain the unique features and difference of various mechanical condition modes, and well meet the requirement for the HMM exemplar training. With the sampled data series from experimental study and onsite measurements, a trained HMM norm modes library was established for different mechanical conditions of the OLTC. A large amount of function verifications demonstrated that the proposed HMM-based mechanical fault diagnosis scheme for OLTC is feasible and effective, with outstanding behavior for fault classification plus an identification rate of 95% in accuracy. An Internet-based program with preferable expandability has also been developed for practical applications of the proposed strategy in HV substations.