Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Machine Learning for Fairness-Aware Load Shedding: A Real-Time Solution via Identifying Binding Constraints

Yuqi Zhou, Joseph Severino, Sanjana Vijayshankar, Juliette Ugirumurera, Jibo Sanyal

TL;DR

This work tackles real-time fairness-aware load shedding by embedding fairness constraints into a DC power-flow based optimization and then leveraging learning to identify the active (binding) constraints. By predicting the binding set with a deep neural network, the original QP is reduced to an equality-constrained problem whose KKT conditions form a sparse linear system, enabling millisecond-scale solutions. The approach is validated on both a 3-bus toy system and the RTS-GMLC 73-bus network, achieving near-exact results and substantial online speedups (up to ~2e4x) while maintaining high prediction accuracy (≈99.5%). The method supports risk-averse loading via CVaR and demonstrates strong potential for real-time, fairness-aware grid operations in practical settings.

Abstract

Timely and effective load shedding in power systems is critical for maintaining supply-demand balance and preventing cascading blackouts. To eliminate load shedding bias against specific regions in the system, optimization-based methods are uniquely positioned to help balance between economic and fairness considerations. However, the resulting optimization problem involves complex constraints, which can be time-consuming to solve and thus cannot meet the real-time requirements of load shedding. To tackle this challenge, in this paper we present an efficient machine learning algorithm to enable millisecond-level computation for the optimization-based load shedding problem. Numerical studies on both a 3-bus toy example and a realistic RTS-GMLC system have demonstrated the validity and efficiency of the proposed algorithm for delivering fairness-aware and real-time load shedding decisions.

Machine Learning for Fairness-Aware Load Shedding: A Real-Time Solution via Identifying Binding Constraints

TL;DR

This work tackles real-time fairness-aware load shedding by embedding fairness constraints into a DC power-flow based optimization and then leveraging learning to identify the active (binding) constraints. By predicting the binding set with a deep neural network, the original QP is reduced to an equality-constrained problem whose KKT conditions form a sparse linear system, enabling millisecond-scale solutions. The approach is validated on both a 3-bus toy system and the RTS-GMLC 73-bus network, achieving near-exact results and substantial online speedups (up to ~2e4x) while maintaining high prediction accuracy (≈99.5%). The method supports risk-averse loading via CVaR and demonstrates strong potential for real-time, fairness-aware grid operations in practical settings.

Abstract

Timely and effective load shedding in power systems is critical for maintaining supply-demand balance and preventing cascading blackouts. To eliminate load shedding bias against specific regions in the system, optimization-based methods are uniquely positioned to help balance between economic and fairness considerations. However, the resulting optimization problem involves complex constraints, which can be time-consuming to solve and thus cannot meet the real-time requirements of load shedding. To tackle this challenge, in this paper we present an efficient machine learning algorithm to enable millisecond-level computation for the optimization-based load shedding problem. Numerical studies on both a 3-bus toy example and a realistic RTS-GMLC system have demonstrated the validity and efficiency of the proposed algorithm for delivering fairness-aware and real-time load shedding decisions.
Paper Structure (7 sections, 1 theorem, 21 equations, 4 figures, 1 table)

This paper contains 7 sections, 1 theorem, 21 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Fairness-Aware Load Shedding Problem
  3. Learning for Real-time Load Shedding
  4. Numerical Simulation
  5. Illustrative Example with a 3-Bus System
  6. RTS-GMLC System
  7. Conclusions and Future Work

Key Result

Corollary 1

Given that the original load shedding problem eq:original_QP has a unique optimal solution, the linear system eq:kkt is solvable and the KKT matrix is nonsingular.

Figures (4)

  • Figure 1: The overall framework for learning the fairness-aware load shedding problem to enable real-time decision-making.
  • Figure 2: A 3-bus illustrative example with two generators.
  • Figure 3: Network topology of the RTS-GMLC system.
  • Figure 4: Load shedding solutions under varying $\delta$ values.

Theorems & Definitions (1)

  • Corollary 1