Expert-guided optimization for load transfer in distribution networks assisted by virtual power plants

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Power Systems

Background:

• Complex distribution networks and distributed energy resources (DERs) challenge reliable post-fault service restoration.
• Integrating DERs for carbon neutrality goals increases intermittency and uncertainty, complicating load transfer decisions.

Purpose of the Study:

• To propose an expert-guided, virtual power plant (VPP)-assisted load transfer optimization framework.
• To improve the speed and reliability of post-fault service restoration in complex power grids with high DER penetration.

Main Methods:

• Developed a topology-aware graph neural network (GNN) for state representation, modeling buses as nodes and switches as edges.
• Implemented a hierarchical reinforcement learning architecture with expert guidance for coordinated network reconfiguration and DER regulation.
• Designed specialized agents for switch operations and DER power adjustments to ensure power balance and voltage security.

Main Results:

• The proposed method achieved faster service restoration and higher load recovery ratios compared to conventional approaches.
• Demonstrated significantly fewer voltage violation events, indicating improved operational safety.
• Showcased enhanced scheduling stability in distribution networks with high DER penetration.

Conclusions:

• The expert-guided, VPP-assisted framework effectively addresses challenges in post-fault restoration for modern power grids.
• Hierarchical graph reinforcement learning provides a robust solution for optimizing load transfer and ensuring grid stability.
• The approach offers significant improvements in efficiency, safety, and reliability for power distribution systems.