Electrical Network Voltage Estimation Using Graph Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Low-voltage electricity distribution networks face challenges with increased integration of decentralized electricity production and new uses like EV charging, leading to voltage level constraints and limited real-time observability due to communication constraints.
Innovation Solution
A method utilizing a graph neural network to estimate voltage amplitudes across nodes in an electrical network, leveraging a small number of measurements, and configuring the network based on topology and electrotechnical characteristics for real-time voltage management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If voltage measurements are collected from all nodes in real-time, then complete observability of the electrical network is achieved, but communication constraints and data transmission burden increase significantly
Solution Approach 1:
The patent divides the electrical network into multiple zones or regions, and selects representative measurement nodes within each zone. Instead of measuring all nodes, the network is segmented into manageable parts where measurements from selected nodes suffice to infer the state of entire zones, reducing communication burden while maintaining observability.
Solution Approach 2:
The patent introduces a state estimation system that acts as an intermediary between physical measurements and network state knowledge. This intermediary uses measurements from a subset of nodes, combined with network topology and electrical laws, to infer the state of unmeasured nodes, thereby reducing the number of required measurements while maintaining complete observability.
2Adaptability or versatility
If decentralized electricity production and EV charging are integrated into LV networks, then network functionality and service capability are improved, but voltage level control and network stability deteriorate
Solution Approach 1:
The patent performs preliminary configuration of the graph neural network with network topology and electrical characteristics before deployment. This preliminary setup enables the system to quickly adapt to varying generation and load patterns from decentralized sources and EV charging, maintaining voltage control reliability without requiring real-time reconfiguration.
Solution Approach 2:
The patent uses graph neural networks to dynamically adjust estimation parameters and model characteristics based on network conditions. As decentralized production and EV charging alter network parameters, the GNN adapts its internal representations and computation to maintain accurate state estimation and voltage control under changing conditions.
3Loss of information
If a large number of nodes are monitored in extensive electrical networks, then comprehensive network management is achieved, but system complexity and computational requirements increase
Solution Approach 1:
The patent creates a virtual copy of the electrical network in the form of a graph structure that mirrors the physical topology. This graphical representation allows the system to model and analyze the entire network state using simplified mathematical relationships, avoiding the need for complex physical measurement and control infrastructure at every node while maintaining comprehensive monitoring capability.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A computer-implemented method of managing an electrical network comprising a set of nodes is provided, which comprises: obtaining voltage measurements for each node of a subset of the set of nodes of the electrical network; configuring a graph neural network based on data representative of a topology of the electrical network; and determining, by the configured graph neural network, an estimate of the respective voltage magnitudes of the nodes of the set of nodes based on the voltage measurements of the nodes of the subset.