| Abstract [eng] |
Every day, electricity networks are subjected to electrical, mechanical, thermal and atmospheric loads, which eventually lead to insulation defects. As insulation deteriorates, partial discharges begin to form and accelerate the deterioration of the insulation. This master‘s thesis deals with partial discharges, their nature, types and detection methods. In order to understand their frequency characteristics, an experimental study was carried out in which fabricated samples with simulated defects were placed in a container of oil and subjected to a high test voltage. The aquired characteristics were used to train a convolutional neural network to determine whether it would be able to make correct predictions of the type of insulation defect based on the measured frequency characteristics. |
