Graph neural networks (GNNs) have been widely studied to handle graph-structured data due to their superior learning capability. Despite the successful applications of GNNs in many areas, their performance suffers heavily from the imbalanced degree distribution (long-tail issue). Most prior studies tackle this issue by graph augmentation, which explicitly increases the communication among nodes by optimizing original topology. In this paper, we employed the perspective of taylor interaction to explore the long-tail issue, and analyzed that there is insufficient interaction between low-degree nodes and their neighbors. In detail, we proposed a novel GNN framework named with degree-aware feature interaction (GraphDAFI), in order to bridge the gap of neighborhood aggregation between head-node embeddings and tail-node embeddings. GraphDAFI comprises two collaborative modules: adaptive feature interaction and degree-aware neighborhood transfer. Adaptive feature interaction leverages node embeddings of the current layer and interactions of the historical layer to perceive potential local information. Then, a unified feature encoder was designed that enhances the interaction to increase the model's generalization ability. To inject relevant information into low-degree nodes, a degree-aware neighborhood transfer was developed, which updates the node-edge adjacency matrix through a degree-aware strategy to achieve knowledge transfer. Experimental results demonstrate that GraphDAFI achieves excellent performance in semi-supervised node classification compared with the state-of-the-art models.
- Article type
- Year
- Co-author
Open Access
Research Article
Issue
To address the challenges of preventing non-random multiple concurrent faults caused by cable aging in shipboard power grids through preventive reconfiguration, and to resolve the issue of unreasonable weight coefficient settings in multi-objective reconfiguration models, thereby enhancing the safety and reconfiguration efficiency of shipboard power grids, a predictive fault reconfiguration method for shipboard power grids based on a double-level optimization strategy is proposed.
A cable aging fault prediction model for shipboard grids was constructed based on Markov chains and thermo-electro-mechanical multi physics analysis. This model was integrated as a constraint into the reconfiguration framework to avoid high-risk branches. A dual-layer optimization strategy was proposed: the upper layer dynamically solves multi-objective weight coefficients using the whale migration algorithm (WMA), while the lower layer determines the optimal switch configuration for grid reconfiguration using a multi-strategy-improved dung beetle optimizer (MSDBO).
After integrating the fault prediction model, the reconfiguration strategy achieved 100% avoidance of high-risk branches (fault probability ≥0.5) proactively. Compared to the conventional two-step passive reconfiguration strategy, convergence speed improved by 47.06%. The dual-layer optimization framework enabled adaptive dynamic adjustment of weight coefficients and increased reconfiguration convergence speed by 56.25%.
The integration of the cable aging fault prediction model and the dual-layer optimization framework effectively enables predictive reconfiguration of shipboard power grids. This approach proactively mitigates non-random faults while significantly improving reconfiguration efficiency and rationality. It offers a novel solution for addressing predictive reconfiguration challenges in non-random multiple-fault scenarios.
京公网安备11010802044758号