Resumo:
Power transformers are one of the most important and expensive components of an electrical
system. Faults in these devices can directly compromise the efficiency and reliability
of the circuit to which they belong. Analysis and detection of incident faults, especially winding
faults, which account for more than 50% of occurrences, are crucial to ensuring reliable
operation of transformers and the electrical sector. Therefore, this dissertation presents a
literature review on these devices, their main components and faults, a review and validation
of the main methods for detecting turns-to-turn and phase-to-ground faults. In addition, a
review of the main simulation models reported in the literature is presented and a computational
model in the ATPDraw environment is proposed that allows modeling of these devices
under the incidence of winding faults, based on parameters determined through electrical
tests or derived from their physical dimensions. The modeling was performed based on the
results of open-circuit and short-circuit tests, and the fidelity of the model to reality was verified
by comparing the simulated results with electrical tests performed on transformers in
which winding faults could be introduced. For detection methods, frequency response testing,
transformation ratio testing, and impulse voltage testing were used. These were performed
on transformers with simulated faults of different intensities, and the ability of these tests to
highlight the occurrence of faults was verified. The results indicate a high efficiency of the
frequency response test in detecting faults of different intensities and transformers of different
construction philosophies. However, for the transformation ratio and impulse voltage
tests, there are conditions that limit their efficiency and restrict their widespread application.
In addition, the proposed model showed a high degree of proximity to the real results
but presented difficulties in representing equipment whose windings are divided into disks.
This work therefore contributes to the state of the art in the analysis and early detection
of faults in power transformer windings, provides a modeling method capable of simulating
transformers with turns-to-turn and phase-to-ground faults, and validates certain methods
for detecting these occurrences. Furthermore, the dissertation exposes the main impacts of
winding disk division on the results of the tests considered and the limitations that this
constructive philosophy entails.
