Time: 4:30 pm
Venue: Room FYW-3316, City University of Hong Kong
Zoom Meeting ID: 899 0158 1173
Password: 123456
Speaker: Miss Biwei Li, City University of Hong Kong
Abstract:The Cascade Risk Mitigation Set (CRMS) of branches in a power network refers to specific combinations of branches that enhance the capacity to mitigate cascading failure risk. However, identifying the optimal combination is computationally challenging due to the need for repeated cascading failure simulations and the combinatorial explosion of possible solutions. In this talk, I will introduce a novel framework based on graph neural networks (GNNs) to efficiently identify optimal CRMS solutions. The identification problem involves the risk assessment of cascading failure in the objective function, which can be naturally cast as a graph regression task and effectively addressed using GNNs. Two key techniques are incorporated to improve the prediction performance of GNNs. First, the Bus and Branch Graph Attention Network (BB-GAT) is proposed to jointly encode bus and branch features into the graph representation. Second, a pre-training strategy is adopted to reduce the dependency on large amounts of task-specific data and to improve generalization across different operational states. Finally, a hybrid approach is designed to identify the optimal CRMS. This approach prunes the search space using a statistical method, then employs BB-GAT to recommend high-potential candidate solutions and ultimately determines the optimal one via cascading failure simulations. Experimental results on the IEEE 39-bus and 118-bus systems show that this framework achieves near-optimal performance while significantly improving computational efficiency over traditional algorithms.
Speaker’s Bio:Biwei Li is currently a Ph.D. candidate in the Department of Electrical Engineering at the City University of Hong Kong, Hong Kong. She received the B.Eng. degree in mechanical engineering from China University of Geosciences, Wuhan, China, in 2015, and the M.Eng. degree in control science and engineering from Beihang University, Beijing, China, in 2018. Her research interests include complex networks, deep learning applications, and robust analysis of power networks.
WEBINAR WEBSITE:
