End-to-End Deep Graph Convolutional Neural Network Approach for Intentional Islanding in Power Systems Considering Load-Generation Balance

Intentional islanding is a corrective procedure that aims to protect the stability of power system during an emergency, by dividing grid into several partitions and isolating elements would cause cascading failures. This paper proposes deep learning method solve problem intentional in end-to-end manner. Two types loss functions are examined for graph partitioning task, function added on model, aiming minimise load-generation imbalance formed islands. In addition, proposed solution incorporates technique merging independent buses their nearest neighbour case there isolated after clusterisation, improving final result cases large complex systems. Several experiments demonstrate introduced provides effective clustering results islanding, managing keep low creating stable Finally, dynamic, relying real-time conditions calculate result.