Distribution Grid Topology Identification With GNN Edge Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for power distribution grid topology identification are complex, challenging to scale, and prone to errors due to incomplete knowledge of switching device statuses, leading to incorrect network topologies and operational inefficiencies.
Innovation Solution
A scalable methodology using graph representations and graph neural networks to combine measurement signals with known topological information, processing edges with different weights to distinguish certain and uncertain connections, enabling accurate node classification and topology determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional topology identification methods are used, then complete knowledge of switching device statuses is required, but the system complexity and computational requirements increase significantly
Solution Approach 1:
The patent introduces graph neural networks as an intermediary between measurement data and topology identification. The GNN processes measurement signals through learned representations, mediating the complex relationship between incomplete switching device status knowledge and accurate topology identification, thereby reducing system complexity while maintaining precision
Solution Approach 2:
The patent transforms the topology identification problem by changing parameters from direct switching device status knowledge to learned representations from measurement signals. The graph neural network learns optimal parameter transformations that enable accurate topology identification without requiring complete knowledge of switching device statuses
2Productivity
If measurement data alone is used for topology identification, then computational scalability is limited, but incorporating known topological information increases processing complexity
Solution Approach 1:
The patent merges measurement signals with known topological information in a unified graph representation framework. The graph neural network processes both data types together, combining their complementary strengths to achieve computational scalability while managing processing complexity through integrated learning
Solution Approach 2:
The graph neural network serves multiple functions simultaneously: it processes measurement signals, incorporates known topological information, handles uncertain connections, and performs topology identification. This multi-functionality improves computational scalability by consolidating multiple processing tasks into a single scalable framework
3Speed
If uncertain connections are treated as known connections, then topology identification speed increases, but prediction accuracy decreases
Solution Approach 1:
The patent applies local quality by treating certain and uncertain connections differently in the graph representation. Certain connections are encoded with high confidence while uncertain connections are encoded with lower confidence, allowing the graph neural network to process them with appropriate weights, thereby maintaining both speed and accuracy
Data Source
AI summary
A computer-implemented method for identifying a topology of a power distribution grid having a number of transformers includes acquiring measurement signals of one or more electrical quantities pertaining to nodes of the power distribution grid. A graph representation is generated using the measurement signals and grid topological information, wherein the measurement signals pertaining to respective nodes are used to derive node features and the grid topological information is used to encode edges representing certain and uncertain connections between the nodes. The graph representation is processed using a graph neural network to classify the nodes and output a mapping of each of the nodes to one of the transformers, whereby a status of the uncertain connections is determined.