Phantom Metering for Unbalanced Grid Fault Diagnostics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing monitoring and diagnostics systems for energy systems, particularly electrical grids, face challenges due to the unbalanced nature of distribution systems, scarcity of measurement units, and inadequacy of current estimation techniques, leading to inaccurate state estimates and unsuitable positive sequence models.
Innovation Solution
A data-driven system utilizing machine learning-based phantom metering and diagnostics engines to estimate measurement information and identify conditions of interest, such as open conductor high-impedance faults, by leveraging topology information and historical data, enabling real-time control and corrective actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If state estimation techniques based on positive sequence models are used, then system observability is improved, but accuracy deteriorates due to unbalanced distribution systems
Solution Approach 1:
The patent transforms the state estimation problem from using positive sequence models to using full three-phase models, changing the mathematical parameters from simplified symmetric components to complete unbalanced phase data, thereby maintaining accuracy in unbalanced distribution systems
Solution Approach 2:
The patent creates virtual phantom meters that replicate the functionality of physical measurement devices using machine learning models, generating synthetic measurement data that copies the information that would be obtained from actual meters installed at every location
2Loss of information
If physical measurement units are installed throughout the system, then measurement coverage is improved, but system cost increases
Solution Approach 1:
The patent creates virtual phantom meters that replicate the functionality of physical measurement devices using machine learning models, generating synthetic measurement data that copies the information that would be obtained from actual meters installed at every location
Solution Approach 2:
The patent makes existing physical meters serve multiple functions by using their data to train machine learning models that generate measurements for multiple virtual phantom meter locations, allowing one physical device to provide information for many virtual measurement points
3Measurement precision
If machine learning-based phantom metering is implemented, then diagnostic accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent performs preliminary training of machine learning models during off-peak periods using historical data, so that when real-time diagnostics are needed, the pre-trained models can make predictions with reduced computational burden compared to training models in real-time
Data Source
AI summary
Methods and systems relating to metering and diagnostics of energy or power systems are provided. Physical measurements are often not available at various parts of an energy system, such as a power grid, for example because there is no meter present or the meter is malfunctioning. Accordingly, phantom metering is performed by estimating measurement information in the energy system. The phantom metering may be based on topology information of the energy system or grid. The energy system is then diagnosed based on the phantom metering information to identify a condition of interest, such as a condition posing danger in the system, a malfunctioning device in the system, an anomaly condition, and so on. The energy system may then be controlled based on the identified condition of interest, for example to take preventative or corrective action.