Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DPFAGA-Dynamic Power Flow Analysis and Fault Characteristics: A Graph Attention Neural Network

Tan Le, Van Le

TL;DR

The paper addresses the need for resilient, data-efficient fault diagnosis and dynamic power-flow analysis in smart grids, where labeled data are scarce and new measurements can appear unexpectedly. It introduces a joint framework combining Graph Attention Networks (GAT), Clustering with Adaptive Neighbors (CAN), and a semi-supervised conditional random field (SSCRF) to learn robust node representations while modeling label dependencies on graph-structured grid data. The approach translates ACOPF into machine learning tasks, proposing End-to-end Prediction and Optimal Constraint Prediction, and leverages CAN to construct adaptive graph structures, with SSCRF providing probabilistic dependencies and GAT delivering end-to-end inference. Numerical results on IEEE bus benchmarks demonstrate improved robustness to missing data and reduced computational latency, indicating practical potential for fast, scalable, and cyber-resilient grid operation and self-healing capabilities.

Abstract

We propose the joint graph attention neural network (GAT), clustering with adaptive neighbors (CAN) and probabilistic graphical model for dynamic power flow analysis and fault characteristics. In fact, computational efficiency is the main focus to enhance, whilst we ensure the performance accuracy at the accepted level. Note that Machine Learning (ML) based schemes have a requirement of sufficient labeled data during training, which is not easily satisfied in practical applications. Also, there are unknown data due to new arrived measurements or incompatible smart devices in complex smart grid systems. These problems would be resolved by our proposed GAT based framework, which models the label dependency between the network data and learns object representations such that it could achieve the semi-supervised fault diagnosis. To create the joint label dependency, we develop the graph construction from the raw acquired signals by using CAN. Next, we develop the probabilistic graphical model of Markov random field for graph representation, which supports for the GAT based framework. We then evaluate the proposed framework in the use-case application in smart grid and make a fair comparison to the existing methods.

DPFAGA-Dynamic Power Flow Analysis and Fault Characteristics: A Graph Attention Neural Network

TL;DR

The paper addresses the need for resilient, data-efficient fault diagnosis and dynamic power-flow analysis in smart grids, where labeled data are scarce and new measurements can appear unexpectedly. It introduces a joint framework combining Graph Attention Networks (GAT), Clustering with Adaptive Neighbors (CAN), and a semi-supervised conditional random field (SSCRF) to learn robust node representations while modeling label dependencies on graph-structured grid data. The approach translates ACOPF into machine learning tasks, proposing End-to-end Prediction and Optimal Constraint Prediction, and leverages CAN to construct adaptive graph structures, with SSCRF providing probabilistic dependencies and GAT delivering end-to-end inference. Numerical results on IEEE bus benchmarks demonstrate improved robustness to missing data and reduced computational latency, indicating practical potential for fast, scalable, and cyber-resilient grid operation and self-healing capabilities.

Abstract

We propose the joint graph attention neural network (GAT), clustering with adaptive neighbors (CAN) and probabilistic graphical model for dynamic power flow analysis and fault characteristics. In fact, computational efficiency is the main focus to enhance, whilst we ensure the performance accuracy at the accepted level. Note that Machine Learning (ML) based schemes have a requirement of sufficient labeled data during training, which is not easily satisfied in practical applications. Also, there are unknown data due to new arrived measurements or incompatible smart devices in complex smart grid systems. These problems would be resolved by our proposed GAT based framework, which models the label dependency between the network data and learns object representations such that it could achieve the semi-supervised fault diagnosis. To create the joint label dependency, we develop the graph construction from the raw acquired signals by using CAN. Next, we develop the probabilistic graphical model of Markov random field for graph representation, which supports for the GAT based framework. We then evaluate the proposed framework in the use-case application in smart grid and make a fair comparison to the existing methods.

Paper Structure

This paper contains 13 sections, 3 equations, 11 figures.

Table of Contents

  1. Introduction
  2. Our Proposed Mechanisms
  3. Problem Formulation and Graph-Based Architecture Solutions
  4. Problem Formulation
  5. Fully Connected Neural Networks
  6. Graph Neural Networks and Graph Attention Networks
  7. Graph Neural Networks and Graph Attention Networks based Framework for Visualization, Prediction, Detection and Classification
  8. Constructing Graph by CAN
  9. Details of Implementation
  10. Numerical Results and Discussion on Applications
  11. Conclusion and Future Directions
  12. Conclusion
  13. Future Directions

Figures (11)

  • Figure 1: Message passing operation at node 1 in GNN.
  • Figure 2: Structure of GNN.
  • Figure 3: Message Passing in GAT.
  • Figure 4: The use of FCNN with full data train size of 100%. It would be done training with 10,000 epochs.
  • Figure 5: The use of GNN with full data train size of 100%. It would be done training with 2000 epochs.
  • ...and 6 more figures