Automatic detection of frequency ranges of power transformer transfer functions for evaluation by mathematical indicators

Sweep Frequency Response Analysis (SFRA) has gained notorious popularity in recent years, due to its high sensitivity to detect electrical failures or winding deformations in transformers. Due to its high sensitivity to several failures and its good reproducibility, this technique is being widely studied, and several contributions to the evaluation and detection have been proposed, which use mathematical indicators for the comparison of frequency response traces. The differences are normally evaluated using fixed frequency ranges, leading to results that are not always accurate. The application of dynamic ranges to the analysis produce a more robust diagnosis, showing that a suitable definition of the frequency ranges improves the ability of the mathematical indicators to detect failures. A smoothing technique can be applied to the SFRA measurements in order to avoid disturbances in the measuring process at high frequencies. This paper integrates the mentioned features and proposes a new improved automatic methodology for dynamic frequency range detection, mathematical indicators evaluation and the subsequent failure interpretation.