Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Leveraging Multi-Task Learning for Multi-Label Power System Security Assessment

Muhy Eddin Za'ter, Amir Sajad, Bri-Mathias Hodge

TL;DR

This work tackles real-time power system security assessment (PSSA) by reformulating it as a multi-label classification problem solved with a multi-task learning framework. A shared conditional auto-encoder provides a common latent representation to four task-specific decoders, enabling simultaneous assessment of static, voltage, transient, and small-signal stability. Using physics-based synthetic data from the IEEE 68-bus system, the approach demonstrates improved F2-scores and robustness to unseen topologies compared with decision trees, XGBoost, and CBDAC, while offering interpretable insights into the type of insecurity. The proposed methodology reduces overfitting through cross-task regularization and adaptive weighting, and it is flexible to add or remove stability criteria, supporting practical deployment in dynamic grid environments. Limitations include assuming full PMU observability and not yet scaling to very large grids, pointing to future work in uncertainty quantification and partial observability.

Abstract

This paper introduces a novel approach to the power system security assessment using Multi-Task Learning (MTL), and reformulating the problem as a multi-label classification task. The proposed MTL framework simultaneously assesses static, voltage, transient, and small-signal stability, improving both accuracy and interpretability with respect to the most state of the art machine learning methods. It consists of a shared encoder and multiple decoders, enabling knowledge transfer between stability tasks. Experiments on the IEEE 68-bus system demonstrate a measurable superior performance of the proposed method compared to the extant state-of-the-art approaches.

Leveraging Multi-Task Learning for Multi-Label Power System Security Assessment

TL;DR

This work tackles real-time power system security assessment (PSSA) by reformulating it as a multi-label classification problem solved with a multi-task learning framework. A shared conditional auto-encoder provides a common latent representation to four task-specific decoders, enabling simultaneous assessment of static, voltage, transient, and small-signal stability. Using physics-based synthetic data from the IEEE 68-bus system, the approach demonstrates improved F2-scores and robustness to unseen topologies compared with decision trees, XGBoost, and CBDAC, while offering interpretable insights into the type of insecurity. The proposed methodology reduces overfitting through cross-task regularization and adaptive weighting, and it is flexible to add or remove stability criteria, supporting practical deployment in dynamic grid environments. Limitations include assuming full PMU observability and not yet scaling to very large grids, pointing to future work in uncertainty quantification and partial observability.

Abstract

This paper introduces a novel approach to the power system security assessment using Multi-Task Learning (MTL), and reformulating the problem as a multi-label classification task. The proposed MTL framework simultaneously assesses static, voltage, transient, and small-signal stability, improving both accuracy and interpretability with respect to the most state of the art machine learning methods. It consists of a shared encoder and multiple decoders, enabling knowledge transfer between stability tasks. Experiments on the IEEE 68-bus system demonstrate a measurable superior performance of the proposed method compared to the extant state-of-the-art approaches.
Paper Structure (28 sections, 17 equations, 5 figures, 2 tables, 1 algorithm)

This paper contains 28 sections, 17 equations, 5 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Proposed Methodology
  3. PSSA as Multi-Label Classification
  4. Architecture
  5. Mathematical Formulation
  6. Input Processing
  7. Shared Conditional Auto-encoder
  8. Task-Specific Decoders
  9. Loss Function
  10. Weight Selection
  11. Training Process
  12. Regularization and Dropout
  13. Advantages of the MTL Approach
  14. Implementation Setup
  15. Database Construction
  16. ...and 13 more sections

Figures (5)

  • Figure 1: High-level overview of the proposed MTL-based PSSA framework.
  • Figure 2: Encoder-decoder architecture of Multi-task Learning for PSSA
  • Figure 3: F2-score (weighted mean of precision and recall) for the compared methods on all security criteria on the IEEE 68 Bus system.
  • Figure 4: Performance comparison between MTL and a single-task encoder-decoder.
  • Figure 5: F2-score comparison when one topology is excluded from training.