Differentiable Optimization for Deep Learning-Enhanced DC Approximation of AC Optimal Power Flow

2025-03-22

arXiv.org

AI Summary

In power systems with high renewable energy penetration, balancing solution accuracy and real-time performance in AC optimal power flow (AC-OPF) remains challenging. Method: This paper proposes a differentiable neural correction framework that jointly integrates differentiable optimization and graph neural networks (GNNs) to learn nonlinear correction parameters for nodal conductances and branch susceptances—enabling the DC-OPF model to implicitly approximate AC physics. By leveraging the implicit function theorem, AC power flow constraints are rendered differentiable; a power flow matching–oriented loss function is designed, and both equality and inequality constraints from AC-OPF are embedded into training. Results: Evaluated on multiple standard test systems, the method reduces DC-OPF active/reactive power errors by over 60%, achieves power flow accuracy comparable to full AC-OPF, and retains millisecond-scale solving time—significantly enhancing both accuracy and computational efficiency for large-scale real-time grid optimization.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Machine Learning: Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

The growing scale of power systems and the increasing uncertainty introduced by renewable energy sources necessitates novel optimization techniques that are significantly faster and more accurate than existing methods. The AC Optimal Power Flow (AC-OPF) problem, a core component of power grid optimization, is often approximated using linearized DC Optimal Power Flow (DC-OPF) models for computational tractability, albeit at the cost of suboptimal and inefficient decisions. To address these limitations, we propose a novel deep learning-based framework for network equivalency that enhances DC-OPF to more closely mimic the behavior of AC-OPF. The approach utilizes recent advances in differentiable optimization, incorporating a neural network trained to predict adjusted nodal shunt conductances and branch susceptances in order to account for nonlinear power flow behavior. The model can be trained end-to-end using modern deep learning frameworks by leveraging the implicit function theorem. Results demonstrate the framework's ability to significantly improve prediction accuracy, paving the way for more reliable and efficient power systems.

Problem

• Enhances DC-OPF to better approximate AC-OPF using deep learning
• Predicts adjusted network parameters to account for nonlinear power flow
• Improves accuracy of power grid optimization under renewable uncertainty

Innovation

• Deep learning framework enhances DC-OPF to mimic AC-OPF
• Neural network predicts adjusted nodal conductances and branch susceptances
• Differentiable optimization enables end-to-end training via implicit function theorem