GIS Partial Discharge Pattern Recognition Based on a Novel Convolutional Neural Networks and Long Short-Term Memory

Area of Science:

• Electrical Engineering
• Artificial Intelligence

Background:

• Partial discharge (PD) detection in gas-insulated switchgear (GIS) is crucial for power system reliability.
• Accurate classification of PD types caused by insulation defects remains a significant challenge.

Purpose of the Study:

• To develop an advanced model for improved PD pattern recognition in GIS.
• To enhance the accuracy of identifying different types of PD signals.

Main Methods:

• A hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) model was developed.
• CNN was used to extract spatial features, while LSTM processed temporal dependencies in PD signals.
• The CNN-LSTM model's performance was compared against traditional analysis methods.

Main Results:

• The CNN-LSTM model achieved the highest pattern recognition rate, averaging 97.9%.
• The proposed model demonstrated superior overall accuracy compared to conventional techniques.
• The model effectively extracts and utilizes spatiotemporal characteristics of PD signals.

Conclusions:

• The CNN-LSTM model offers a reliable and accurate solution for PD diagnosis in GIS.
• This approach significantly improves the accuracy of PD signal pattern recognition.
• The findings provide a valuable reference for enhancing the safety and maintenance of GIS equipment.