| Summary: | Recognizing partial discharge (PD) patterns is important for identifying insulation breakdown in high-voltage (HV) systems, and recent improvements in deep learning algorithms provide new prospects for improving PD event detection and classification. Early and accurate detection of PD patterns is critical for avoiding equipment breakdowns and ensuring the reliability of electrical systems. However, traditional PD detection approaches frequently have disadvantages, such as issues capturing confusing internal discharges, prolonged manual processing, and the exposure to human interpretation errors. The research aims to predict PD pattern using Deep Learning (DL) methods with different types of PD such as corona, surface etc. The methodology was divided into three step procedure of PD experimental set-up. First step is the different of PD defect such as corona, surface and internal are used for this investigation. Second step is experimental setup which consists of PD box and PD measurements. For the DL analysis, MATLAB Toolbox was utilized. Final setup where a large dataset of PD patterns is gathered and examined to train and test deep learning models, and various graphical analyses are used to evaluate classification accuracy and strength across different PD types. The results obtained will show that deep learning approaches may considerably increase the accuracy, efficiency, and reliability of PD pattern identification, thus providing an attractive option for improving future maintenance strategies in high-voltage electrical equipment.
|
|---|
