Fault detection in electrical power systems using attention-GRU-based fault classifier (AGFC-Net)

Area of Science:

• Engineering
• Computer Science
• Artificial Intelligence

Background:

• Fault detection is critical for system reliability and productivity.
• Conventional methods struggle with accuracy, feature extraction, and generalizability.
• Unaddressed faults lead to system failures and costly downtimes.

Purpose of the Study:

• To develop an advanced fault detection system overcoming limitations of conventional approaches.
• To enhance feature extraction and temporal dependency learning for improved fault classification.

Main Methods:

• Proposed an Attention-GRU-Based Fault Classifier (AGFC-Net).
• Integrated a sophisticated attention mechanism with Gated Recurrent Units (GRU).
• Focused on key fault features and temporal correlations for classification.

Main Results:

• AGFC-Net achieved a high fault detection accuracy of 99.52%.
• Demonstrated superior performance compared to conventional machine learning and deep learning algorithms.
• Showcased effective performance even in noisy conditions.

Conclusions:

• AGFC-Net offers a robust, adaptive, and scalable solution for autonomous fault diagnosis.
• The method enhances intelligent and trustworthy fault detection systems.
• Paves the way for improved reliability in power grids and industrial applications.