Towards Understanding the Unreasonable Effectiveness of Learning AC-OPF Solutions

TL;DR

This paper investigates why deep neural networks can accurately and robustly approximate optimal power flow solutions, revealing underlying factors and proposing a new model to enhance performance in power systems optimization.

Contribution

It uncovers the reasons behind DNN effectiveness in OPF approximation and introduces a novel model leveraging these insights for improved accuracy and robustness.

Findings

DNN output volatility relates to generator behavior

Characteristics influencing DNN learning are identified

A new robust and accurate OPF prediction model is proposed

Abstract

Optimal Power Flow (OPF) is a fundamental problem in power systems. It is computationally challenging and a recent line of research has proposed the use of Deep Neural Networks (DNNs) to find OPF approximations at vastly reduced runtimes when compared to those obtained by classical optimization methods. While these works show encouraging results in terms of accuracy and runtime, little is known on why these models can predict OPF solutions accurately, as well as about their robustness. This paper provides a step forward to address this knowledge gap. The paper connects the volatility of the outputs of the generators to the ability of a learning model to approximate them, it sheds light on the characteristics affecting the DNN models to learn good predictors, and it proposes a new model that exploits the observations made by this paper to produce accurate and robust OPF predictions.