Distributed Nonlinear State Estimation in Electric Power Systems using Graph Neural Networks

TL;DR

This paper presents a graph neural network approach for nonlinear power system state estimation that is noniterative, scalable, and robust against cyber attacks and communication issues, improving over traditional methods.

Contribution

It introduces a novel GNN-based nonlinear state estimation method that is linear in complexity, distributed, and resilient to cyber and communication disruptions.

Findings

Achieves accurate state estimation with linear inference complexity.

Demonstrates robustness against cyber attacks and communication irregularities.

Outperforms traditional iterative methods in stability and scalability.

Abstract

Nonlinear state estimation (SE), with the goal of estimating complex bus voltages based on all types of measurements available in the power system, is usually solved using the iterative Gauss-Newton method. The nonlinear SE presents some difficulties when considering inputs from both phasor measurement units and supervisory control and data acquisition system. These include numerical instabilities, convergence time depending on the starting point of the iterative method, and the quadratic computational complexity of a single iteration regarding the number of state variables. This paper introduces an original graph neural network based SE implementation over the augmented factor graph of the nonlinear power system SE, capable of incorporating measurements on both branches and buses, as well as both phasor and legacy measurements. The proposed regression model has linear computational complexity during the inference time once trained, with a possibility of distributed implementation. Since the method is noniterative and non-matrix-based, it is resilient to the problems that the Gauss-Newton solver is prone to. Aside from prediction accuracy on the test set, the proposed model demonstrates robustness when simulating cyber attacks and unobservable scenarios due to communication irregularities. In those cases, prediction errors are sustained locally, with no effect on the rest of the power system's results.