🤖 AI Summary
Existing neural state estimation methods for distribution systems lack zero-shot adaptability to unknown topology changes—such as sensor failures or branch switching—limiting their reliability in dynamic operational scenarios.
Method: This paper pioneers the systematic integration of Graph Neural Networks (GNNs) into a zero-shot distribution system state estimation framework. We propose a topology-aware data augmentation strategy and reveal a non-monotonic relationship between GNN depth and robustness—challenging the common “deeper implies more robust” assumption. Furthermore, we combine zero-shot learning mechanisms with grid-search-based hyperparameter optimization to enhance generalization to unseen topologies.
Results: Experiments across multiple zero-shot settings demonstrate an average 32% reduction in estimation error, validating the proposed method’s superior robustness and practical applicability in dynamically reconfigured distribution networks.
📝 Abstract
This paper addresses the challenge of neural state estimation in power distribution systems. We identified a research gap in the current state of the art, which lies in the inability of models to adapt to changes in the power grid, such as loss of sensors and branch switching, in a zero-shot fashion. Based on the literature, we identified graph neural networks as the most promising class of models for this use case. Our experiments confirm their robustness to some grid changes and also show that a deeper network does not always perform better. We propose data augmentations to improve performance and conduct a comprehensive grid search of different model configurations for common zero-shot learning scenarios.