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

TL;DR

This paper introduces an augmented hierarchical graph neural network (AHGNN) to efficiently predict critical lines and generate N-1 reduced optimal power flow solutions, significantly decreasing computation time while maintaining solution accuracy.

Contribution

The paper presents a novel AHGNN model for N-1 OPF that outperforms existing GNN variants in speed and accuracy, advancing machine learning applications in power system optimization.

Findings

AHGNN significantly reduces computation time for N-1 OPF.

The approach maintains high solution quality comparable to traditional methods.

Benchmark models validate the effectiveness of AHGNN over other GNN variants.

Abstract

Optimal power flow (OPF) is used to perform generation redispatch in power system real-time operations. N-1 OPF can ensure safe grid operations under diverse contingency scenarios. For large and intricate power networks with numerous variables and constraints, achieving an optimal solution for real-time N-1 OPF necessitates substantial computational resources. To mitigate this challenge, machine learning (ML) is introduced as an additional tool for predicting congested or heavily loaded lines dynamically. In this paper, an advanced ML model known as the augmented hierarchical graph neural network (AHGNN) was proposed to predict critical congested lines and create N-1 reduced OPF (N-1 ROPF). The proposed AHGNN-enabled N-1 ROPF can result in a remarkable reduction in computing time while retaining the solution quality. Several variations of GNN-based ML models are also implemented as benchmark to demonstrate effectiveness of the proposed AHGNN approach. Case studies prove the proposed AHGNN and the associated N-1 ROPF are highly effective in reducing computation time while preserving solution quality, highlighting the promising potential of ML, particularly GNN in enhancing power system operations.