Time-Frequency Analysis for the Interpretation of FRA Measurements

MOMBELLO, ENRIQUE E.; VACA VARGAS, PAMELA

Berlin

Simposio; VDE-Fachtagung Hochspannungstechnik 2016; 2016

VDE

FrequencyResponse Analysis (FRA) is a powerful and non-destructive test to diagnose theinternal condition of a power transformer. This technique enables the detectionof different kinds of faults related to mechanical and electrical problems thatoccur inside the transformer without the need of an internal inspection. Thisfeature aroused not only the interest of researchers in the field oftransformer diagnosis in the last decade but also has captured the attention oftransformer manufacturers and electrical utilities because of its potential toverify the internal integrity of transformers, detecting incipient faults as apreventive strategy of maintenance or confirming failures after catastrophicevents like lightning, external short circuits, earthquakes. FRA works as acomparative method, i.e. the detection of a damage or abnormal condition isdone by comparing a current set of FRA measurements versus a well-knownreference signature or fingerprint, that is, a measurement performed on thetransformer in sound condition. The main goal is detect differences between FRAtraces which are helpful to infer the condition of the unit.Thestandardization of FRA tests is nowadays incomplete. There are technicalreports, recommendations, standards and user manuals available for thispurpose. These documents have in common some theoretical frameworks anddefinitions but unfortunately they do not give a consolidated methodology forthe interpretation of FRA measurements. Today the industry and customers areseeking for more guidance on the interpretation of FRA results to avoidmisinterpretations. Therefore, there is currently a necessity for thedevelopment of a FRA interpretation algorithm to be applied to conditionassessment. In the last years some diagnostic interpretation methods based onequivalent circuits, parameter identification of transfer functions andstatistical indicators have been proposed, but none of them has been entirelyconclusive.Thisdocument proposes a new approach for FRA interpretation using a novel toolknown as Time-Frequency Analysis (TFA) which has already shown successfulresults in other fields, such as detection of diseases or syndromes usingbio-signals, speaker identification or detection of faults in loudspeakersusing sound signals or machine condition monitoring and diagnosis usingelectrical or vibrational signals.