Resumo:
The transformers are among the most important equipments in the electric system.
Non-scheduled suspensions can bring considerable nuisance and large financial losses. The
only way to avoid the problems caused by flaws in transformers is through a good
maintenance strategy.
A typical flaw that deserves special attention is the partial discharge because its
silent development and capability to cause non-scheduled suspensions in the transformer.
Partial discharge is a problem in the electrical insulation that reveals itself over tiny arcs into
the dielectric material. It degrades the insulation until the complete fault and possible
destruction of the equipment. Partial discharge activity in transformer must be monitored in
the way to keep track its evolution and plan an intervention before the occurrence of a
catastrophic fault.
Among of different measurement methods of partial discharge in transformers, the
acoustic emission technique deserves attention. It captures acoustic signals emitted by
discharges in every location into the transformer. One of its advantages is the ability to locate
the source of partial discharge. However, to feature theses acoustic signals is an important
task of the location methods.
Only detect partial discharge activity into de transformer is not enough to an efficient
maintenance in this equipment. It’s imperative to know the location where the defect is
developing itself to make possible an assessment of the risk and the planning an
intervention.
The objective of this work is to present the development of a partial discharge
localization algorithm using acoustic emissions. This is an initial work where the basic tools
are developed to realize this task. The localization algorithm is divided in two stages: to
extract features of the signals and the localization itself. The feature extraction is
accomplished through the Discrete Wavelet Transform. The localization is performed by a
Genetic Algorithm. Experiments in a test tank simulate partial discharges in insulation oil and
provide signals to analyze its features and to test the localization algorithm.