Self-Supervised Learning of Parametric Approximation for Security-Constrained DC-OPF

TL;DR

SC-DCOPF is computationally intensive due to large contingency sets. We present a self-supervised framework that uses a Graph Attention Network to predict tunable line-capacity scalings $\boldsymbol{\alpha}$ and solves a differentiable, parametric DC-OPF, with an implicit loss that accounts for pre- and post-contingency costs. The approach preserves DC-OPF physics while enabling end-to-end training without labeled SC-DCOPF solutions. Empirical results on IEEE 57-, 118-, and 200-bus systems show high dispatch accuracy, low cost approximation error, and strong data efficiency, outperforming semi-supervised and end-to-end baselines. This yields a scalable, interpretable option for real-time secure power system operations.

Abstract

This paper introduces a self-supervised learning framework for approximating the Security-Constrained DC Optimal Power Flow (SC-DCOPF) problem using a parametric linear model. The approach preserves the physical structure of the DC-OPF while incorporating demand-dependent tunable parameters that scale transmission line limits. These parameters are predicted via a Graph Neural Network and optimized through differentiable layers, enabling direct training from contingency costs without requiring labeled data. The framework integrates pre- and post-contingency optimization layers into an implicit loss function. Numerical experiments on benchmark systems demonstrate that the proposed method achieves high dispatch accuracy, low cost approximation error, and strong data efficiency, outperforming semi-supervised and end-to-end baselines. This scalable and interpretable approach offers a promising solution for real-time secure power system operations.

Figures (2)

• Figure 1: The proposed learning framework.
• Figure 2: Cost correlation with respect to SC-DCOPF model.