Electric Grid Mapping Using Graph Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electric grid models are often incomplete or inaccurate due to obscured power lines by vegetation, buildings, and bridges, and lack representation of underground lines, making it challenging to accurately map connections between electrical nodes and loads.
Innovation Solution
A method involving the creation of a network graph representation of an electric power grid using machine learning, specifically a graph neural network, to predict and verify wire connections by generating an over-connected graph and annotating edges as positive or negative links, thereby updating the model to include missing connections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If overhead or street-level imaging is used to map the grid, then mapping coverage is improved, but obscured power lines by vegetation, buildings, and bridges are not detected
Solution Approach 1:
The patent introduces an intermediary inference mechanism that connects observable grid components (transformers, poles, switches) to unobservable power line connections. The system uses the spatial relationships and operational characteristics of these visible components to infer the presence and configuration of obscured power lines, effectively using intermediaries to bridge the detection gap.
Solution Approach 2:
The patent replaces direct visual/mechanical detection of power lines with a computational inference system. Instead of relying on physical imaging to directly observe power lines, the system substitutes a machine learning model that processes grid component data to predict power line connections, replacing the mechanical imaging approach with an intelligent computational system.
2Measurement precision
If complete grid data is available, then model accuracy is improved, but data from grid operators is difficult to obtain and incomplete
Solution Approach 1:
The system performs self-service by autonomously inferring missing grid connection data from available observations without requiring external data from grid operators. The machine learning model processes observable grid component information to self-determine power line configurations, eliminating the need for manual data collection from operators while achieving complete model accuracy.
Solution Approach 2:
The patent performs preliminary action by pre-establishing the relationships between grid components and power line connections through machine learning training. The system pre-processes available grid data to create a predictive model that can infer missing connections before actual modeling is needed, preparing the inference capability in advance to compensate for incomplete operator data.
3Measurement precision
If machine learning inference is used to predict connections, then missing wire connections are identified, but computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the complex grid modeling task into distinct components: observable component detection, spatial relationship analysis, and connection inference. The machine learning model is segmented to process specific features (transformer locations, pole positions, switch configurations) separately before integrating them for power line prediction, breaking down computational complexity into manageable segments.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for developing electrical grid mapping. One of the methods includes obtaining a computer model of an electric power grid; generating a network graph representation of the computer model, wherein nodes of the network graph represent grid assets of the computer model and edges of the network graph represent wires connecting the grid assets; generating an initial prediction of links between nodes in the network graph by adding at least one edge to the network graph to obtain an over-connected graph; applying the over-connected network graph as input to a machine learning model to obtain an annotated network graph, the machine learning model configured to identify edges as positive links and negative links, and apply annotations to the edges indicating whether each edge is a positive or negative link; and updating the model based on the annotated network graph.