Method for locating secondary circuit faults in substations based on graph neural networks

Area of Science:

• Electrical Engineering
• Computer Science

Background:

• Smart substations rely on secondary circuits for operational control and monitoring.
• Accurate fault location in these circuits is crucial for maintaining grid stability and preventing cascading failures.

Purpose of the Study:

• To develop an accurate and portable fault location method for smart substation secondary circuits.
• To leverage graph neural networks for enhanced fault detection capabilities.

Main Methods:

• Constructed a graph database model of secondary circuits to represent device interconnections.
• Defined fault location as a graph classification problem using a novel fault graph generation model.
• Developed a graph neural network (GNN) based fault location model optimized with binary cross-entropy.

Main Results:

• The proposed GNN method demonstrated higher fault location accuracy compared to existing models.
• Experiments showed improved robustness in identifying faults across varied network topologies and configurations.
• Case studies on a 220 kV smart substation validated the method's effectiveness.

Conclusions:

• The graph neural network approach offers a significant advancement in secondary circuit fault location for smart substations.
• This method provides a robust and accurate solution for improving the reliability of power grids.