Abstract:Optimal power flow (OPF) is a critical optimization problem that allocates power to the generators in order to satisfy the demand at a minimum cost. Solving this problem exactly is computationally infeasible in the general case. In this work, we propose to leverage graph signal processing and machine learning. More specifically, we use a graph neural network to learn a nonlinear parametrization between the power demanded and the corresponding allocation. We learn the solution in an unsupervised manner, minimizing the cost directly. In order to take into account the electrical constraints of the grid, we propose a novel barrier method that is differentiable and works on initially infeasible points. We show through simulations that the use of GNNs in this unsupervised learning context leads to solutions comparable to standard solvers while being computationally efficient and avoiding constraint violations most of the time.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the Optimal Power Flow (OPF) problem in power systems. Specifically, OPF is a crucial optimization problem, aiming to allocate the power of generators at the lowest cost to meet the demands in the power grid. However, due to the non - convex nature of the OPF problem, exact solutions are generally computationally infeasible and belong to NP - hard problems. To overcome this challenge, the paper proposes an unsupervised learning method based on Graph Neural Networks (GNNs). By using GNNs, the authors aim to learn a nonlinear parameterized mapping that can relate the demanded power to the corresponding power allocation. This method can not only solve the problem computationally efficiently but also avoid constraint violations most of the time. The main contributions of the paper include: 1. Using GNNs to parameterize the mapping from the states of each node in the power grid to the target generated power. 2. Learning this mapping by optimizing the constrained OPF problem. 3. Introducing a differentiable piecewise penalty function based on the logarithmic barrier function to enforce the constraints. 4. Proposing new methods for evaluating machine - learning OPF models, considering not only the frequency of model constraint violations but also the severity of the violation degree. 5. Demonstrating the applicability of graph neural networks in the optimal power flow problem, especially the good scalability on sparse graphs and the possibility of distributed implementation. Through these contributions, the paper provides a new and efficient method for solving the OPF problem, which is expected to improve the operational efficiency and reliability of power systems in practical applications.

Unsupervised Optimal Power Flow Using Graph Neural Networks

Optimal Power Flow Using Graph Neural Networks

Proximal Policy Optimization with Graph Neural Networks for Optimal Power Flow

Graph Neural Network-Accelerated Network-Reconfigured Optimal Power Flow

A Physics-Guided Graph Convolution Neural Network for Optimal Power Flow

Topology-aware Graph Neural Networks for Learning Feasible and Adaptive ac-OPF Solutions

AC Optimal Power Flows: Combining Lagrangian Dual and physics-Guided Graph Neutral Networks

Optimal Power Flow Based on Physical-Model-Integrated Neural Network with Worth-Learning Data Generation

Constraints and Variables Reduction for Optimal Power Flow Using Hierarchical Graph Neural Networks with Virtual Node-Splitting

Determining Feasibility of DC Power Flow Using Graph Neural Networks

Unsupervised Learning for Solving AC Optimal Power Flows: Design, Analysis, and Experiment

DeepOPF+: A Deep Neural Network Approach for DC Optimal Power Flow for Ensuring Feasibility

Solving AC Power Flow with Graph Neural Networks under Realistic Constraints

Unsupervised Deep Learning for AC Optimal Power Flow via Lagrangian Duality

OptNet-Embedded Data-Driven Approach for Optimal Power Flow Proxy

N-1 Reduced Optimal Power Flow Using Augmented Hierarchical Graph Neural Network

Optimal Power Flow in Highly Renewable Power System Based on Attention Neural Networks

Graph Convolutional Neural Networks for Optimal Power Flow Locational Marginal Price

Learning to Solve the AC Optimal Power Flow via a Lagrangian Approach

Optimal Power Flow Solutions Incorporating Stochastic Wind and Solar Power

Optimal Power Flow Estimation Using One-Dimensional Convolutional Neural Network