School of
Information Technology and Electrical Engineering

Speaker: Jakob Pallot
Seminar Date: Wed, 29/01/2020 - 11:00
Venue: 78-343
Host: Dr Chandima Ekanayake

Seminar Type:  PhD Confirmation Seminar

Abstract: 

Power transformers are among the most expensive and vital pieces of equipment in an electrical grid with failure of the transformers potentially costing power companies millions of dollars a year in restoration, replacement, and labour costs. One way in which power companies can limit this is with implementing condition-monitoring techniques designed to monitor the state of various parts within the transformer. These components can be the bushings, copper windings, iron core, dielectric solid insulation, oil insulation, tap-changer, pumps, motors, transformer casing, and other supporting parts. These techniques can range from analysis of the moisture present within the oil, the insulation oil breakdown strength, turns ratio, partial discharges, dissipation factor, Frequency Response Analysis (FRA) and recently, mechanical vibrations.

Mechanical vibrations are generated within the transformer primarily from the core and winding, with supplementary vibrations coming from auxiliary components such as pumps, motors, fans, and ambient noise. Analysis of these mechanical vibrations for transformer condition monitoring became of concern for engineers when little was achieved on properly interpreting vibro-acoustic signals measured from an in-operation transformer to assess its winding structure. Changes to the internal condition of the transformer, such as deformations, affected these vibrations whether it be the vibrations during normal operation or the natural resonance (natural frequencies) of the transformer. This gave rise to two different fields of study, steady-state vibration analysis, and modal analysis.

Steady-state vibration analysis focuses on changes to the vibration pattern during normal operation, typically with a constant load, whilst modal analysis aims at detecting changes to the natural frequencies of the structure, the frequencies at which the structure resonates the most. Furthermore, within modal analysis there are two techniques, Experimental Modal Analysis (EMA) and Operational Modal Analysis (OMA), with EMA being the well-established technique yet only applicable in offline cases, and OMA being the newly developed technique but can be applied in a semi-online capacity. Progression in the above two fields for transformer winding condition assessment is relatively in the early stages and further investigations are required.

The aim of this project is to investigate the correlations between the vibration pattern and the winding condition changes such as deformation, bulging, loss of winding clamping pressure, etc. The primary focus of this work is being split between analysing the steady-state vibrations and applying OMA to the structure and improving the two techniques. Given the complex nature of transformer vibrations, work is initially being focussed towards single-phase transformers since the simplified geometry allows for better cohesion between Finite Element Model (FEM) simulations and experimental data. With a strong cohesion between experiments and FEM simulations, better assumptions and inferences can be made from the data, which can then be implemented in the industrial cases of three-phase transformers, potentially solving the confusion dominating this area of research.

Preliminary work of FEM simulations and experiments have shown that due to the existence of mode shapes on the surface of the transformer at frequencies 100 Hz, 200 Hz, 300 Hz and 400 Hz the signal-to-noise ratio of the vibrations exhibited a different pattern per frequency, rather than a uniform SNR pattern across the transformer. However, in current interpretation techniques this scenario has not been addressed. The proposed approach is therefore useful to increase the reliability of current vibration data interpretations when analysing structural changes from the steady-state vibrations of transformers. Additionally, the Frequency Domain Decomposition technique for OMA showed positive results in identifying the natural frequencies of the transformer structure when applying turn-off transient responses.  This allows for the development of a new technique for vibration analysis of field-installed transformers.

Biography: 

Jakob completed a Bachelor of Engineering (Electrical) at Griffith University in 2018 and as part of his undergraduate thesis he was the recipient of the IEEE Best Power Engineering Thesis for the IEEE QLD section. After completing his undergraduate degree he transferred to the University of Queensland to start his PhD under the guidance of his previous undergraduate thesis coordinator, Dr. Chandima Ekanayake. Jakob has a keen interest in improving condition monitoring techniques in a cost-effective manner and is currently working on improving vibro-acoustic monitoring for use in power transformers.