Towards Understanding the Unreasonable Effectiveness of Learning AC-OPF Solutions
This addresses a knowledge gap in power systems for researchers and practitioners, but it is incremental as it builds on existing work without introducing a major breakthrough.
The paper tackles the problem of understanding why deep neural networks can effectively approximate solutions to the Optimal Power Flow (OPF) problem, and it proposes a new model that produces accurate and robust predictions based on insights about generator volatility and model characteristics.
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.