Graph Convolutional Neural Network for DC Microgrid Instability Fault Location
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for locating instability fault sources in direct-current microgrids are limited and lack automated solutions, leading to inefficiencies in maintenance and repair.
Innovation Solution
A graph theory-based method and device that utilizes electrical data from power electronic converters to construct a graph model, which is then processed using a graph convolutional neural network to diagnose instability and identify fault sources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional manual methods are used to locate instability fault sources, then system stability can be maintained, but maintenance efficiency and fault location accuracy deteriorate
Solution Approach 1:
The patent replaces manual analysis methods with automated graph convolutional neural network algorithms. The system automatically constructs graph models from electrical data, performs feature extraction, and identifies fault sources through machine learning, eliminating the need for manual system analysis and significantly improving both efficiency and accuracy.
Solution Approach 2:
The patent creates graph model representations (copies) of the physical microgrid system. By constructing graph models where nodes represent converters and edges represent coupling relationships, the system can analyze fault conditions in the model without disrupting the actual system, enabling accurate fault location through model analysis.
2Extent of automation
If automated solutions are implemented for fault location, then maintenance efficiency improves, but system complexity increases
Solution Approach 1:
The patent segments the complex fault diagnosis task into distinct modular components: graph model construction, feature extraction, graph convolutional neural network processing, and fault identification. Each module handles a specific aspect of the analysis, making the overall automated system more manageable and easier to implement despite the high level of automation.
3Measurement precision
If graph models and neural networks are used to improve fault detection accuracy, then measurement precision improves, but device complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-processing electrical data to extract relevant features before feeding them into the graph convolutional neural network. The system pre-construccts graph models from system topology and pre-processes electrical measurements, reducing the computational burden during actual fault detection and improving accuracy without proportionally increasing complexity.
Data Source
AI summary
Disclosed in the present invention is a graph theory-based method for locating an instability fault source in a direct-current microgrid. The method includes: collecting key electrical feature quantities through an instability fault feature collection and extraction link to accurately capture an instability oscillation phenomenon, and utilizing wavelet packet decomposition to construct fault feature vectors; then, according to the structure of the direct-current microgrid, constructing an undirected graph model G=(V, E, X), which lays a foundation for analysis; applying a graph convolutional neural network for deep learning and training to further improve the accuracy and efficiency of diagnosis; deploying the trained model in an actual system to effectively achieve accurate fault diagnosis and location of an instability source. Also provided in the present invention is a device for locating an instability fault source in a direct-current microgrid. The method provided by the present invention can significantly improve a maintenance level of the direct-current distribution network, and greatly improve a detection and adjustment ability of system operation stability through advanced automatic diagnosis technology.