Techniques for Failures Classification in Power Transformers by High-Frequency Current Transformers

The interest about transformer monitoring systems has significantly increased in recent years. This interest comes both from academia and industry, as the detection of failures provides a high degree of control over the operating conditions of electrical machines and significant cost savings through timely maintenance. In the stages that precede the total failure of a transformer, it is common to detect partial or full discharges that initiate due to degradation of the dielectric materials that form the insulation in the transformer. Therefore, the identification and classification of these phenomena can indicate the quality of the insulation and allow complete failure of the electrical machine to be avoided. One of the main points to consider is where the fault is occurring. For example, a fault can occur in both the active part of the transformer and in its insulators. Therefore, diagnosing the type of failure is very important as they require different maintenance actions. The frequency spectrum analysis is one of the mostly effective and applied methods in the area of signal processing, because it is possible to extract metrics and observe singular characteristics in a signal One of the principal non-invasive methods is current pulse analysis is using a high frequency current sensor, known as High-Frequency Current Transformer (HFCT), with a bandwidth of up to 50 MHz, to evaluate differences in the current signals originating from discharges in insulators from the total discharges that occur in the active part of an oil-isolated transformer. From this perspective, the Fourier Transform was applied to the acquired current signals and some parameters were extracted, like average bandwidth, skewness, and kurtosis to determine which of these combinations is the most effective in classifying the verified failures.