Machine Learning for Solving Optimal Power Flow Problems

Speaker: Minghua Chen

Venue: 418, Teaching Building No.4, Yuquan Campus

Abstract: The Optimal Power Flow (OPF) problem aims to minimize power generation costs while meeting grid demand and adhering to physical constraints, serving as a pivotal challenge in power system operations, its applications span electricity markets, reliability assessments, and grid modernization for carbon neutrality, with global annual operational costs reaching billions of USD. Traditional iterative solvers struggle with large-scale OPF due to its non-convex and NP hard nature.Recent advances in machine learning (ML)have accelerated OPF solutions by orders of magnitude through techniques like constraint classification, warm-start solutions, and direct load-to-solution mapping, while maintaining accuracy.This tutorial explores cutting-edge ML applications in OPF, including active constraint identification, solver strategy learning, and cross-domain optimization generalization.Future directions for enhancing computational efficiency and robustness under increasing renewable energy uncertainty are also discussed.