Dual Conic Proxies for AC Optimal Power Flow

Guancheng Qiu,Mathieu Tanneau,Pascal Van Hentenryck

2024-03-26

Abstract:In recent years, there has been significant interest in the development of machine learning-based optimization proxies for AC Optimal Power Flow (AC-OPF). Although significant progress has been achieved in predicting high-quality primal solutions, no existing learning-based approach can provide valid dual bounds for AC-OPF. This paper addresses this gap by training optimization proxies for a convex relaxation of AC-OPF. Namely, the paper considers a second-order cone (SOC) relaxation of AC-OPF, and proposes \revision{a novel architecture} that embeds a fast, differentiable (dual) feasibility recovery, thus providing valid dual bounds. The paper combines this new architecture with a self-supervised learning scheme, which alleviates the need for costly training data generation. Extensive numerical experiments on medium- and large-scale power grids demonstrate the efficiency and scalability of the proposed methodology.

Machine Learning,Optimization and Control

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: although existing machine - learning optimization proxies can provide high - quality primal solutions when predicting the AC Optimal Power Flow (AC - OPF) problem, they are unable to provide effective dual bounds. This results in these proxies lacking optimality certificates, which restricts their application in actual power systems. Therefore, this paper aims to fill this gap by training optimization proxies for convex - relaxation AC - OPF, especially using second - order cone (SOC) relaxation, and proposes a new architecture that embeds a fast, differentiable dual - feasibility recovery mechanism, thereby providing effective dual bounds.

Specifically, the main contributions of the paper include:
1. Proposing a dual - optimization proxy that utilizes the convex relaxation of AC - OPF to obtain effective dual bounds.
2. Developing a new dual - feasible architecture - Dual Conic Proxies (DCP), which utilizes conic duality to ensure dual feasibility.
3. Adopting a self - supervised training algorithm to train DCP, thus eliminating the need for expensive and potentially numerically unstable optimization solvers.
4. Conducting numerical experiments on large - scale power systems, demonstrating the scalability and high - efficiency of the DCP method and its ability to provide high - quality certified dual bounds.

By solving these problems, the paper provides the first learning - based effective dual - bound prediction method for the AC - OPF problem, which is of great significance for the economic operation of power systems.