Detection of Corona Faults in Switchgear by Using 1D-CNN, LSTM, and 1D-CNN-LSTM Methods.

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Corona faults are a significant concern in metal-clad switchgear, leading to flashovers and potential equipment damage.
• Electrical stress and poor air quality are primary causes of corona faults, necessitating effective detection methods.
• Deep Learning (DL) offers autonomous feature learning for improved fault detection in electrical equipment.

Purpose of the Study:

• To systematically analyze and compare three DL models (1D-CNN, LSTM, 1D-CNN-LSTM) for detecting corona faults in switchgear.
• To identify the most effective DL model for accurate corona fault detection.
• To evaluate model performance in both time and frequency domains.

Main Methods:

• Analysis of three deep learning techniques: 1D-CNN, LSTM, and a hybrid 1D-CNN-LSTM model.
• Utilizing sound wave data generated within switchgear for fault detection.
• Performance evaluation in both time domain analysis (TDA) and frequency domain analysis (FDA).

Main Results:

• The hybrid 1D-CNN-LSTM model demonstrated superior performance compared to individual 1D-CNN and LSTM models.
• In TDA, 1D-CNN-LSTM achieved success rates of 99.3%, 98.4%, and 98.4% for training, validation, and testing.
• In FDA, 1D-CNN-LSTM achieved perfect 100% success rates across training, validation, and testing.

Conclusions:

• The hybrid 1D-CNN-LSTM model is the most effective for detecting corona faults in switchgear.
• The model's high accuracy in both TDA and FDA makes it a robust solution for fault identification.
• Accurate corona fault detection is critical for preventing flashovers and ensuring operational safety in electrical equipment.