Data-Driven Flow and Injection Estimation in PMU-Unobservable Transmission Systems

TL;DR

This paper introduces a machine learning approach using deep neural networks to accurately estimate power flows and injections in power systems with limited PMU measurements, outperforming existing data-driven methods.

Contribution

It presents a novel DNN-based method that incorporates physical constraints for improved flow and injection estimation in sparsely observed power systems.

Findings

More accurate estimation in unobservable systems

Outperforms other data-driven methods

Effective in IEEE 118-bus system

Abstract

Fast and accurate knowledge of power flows and power injections is needed for a variety of applications in the electric grid. Phasor measurement units (PMUs) can be used to directly compute them at high speeds; however, a large number of PMUs will be needed for computing all the flows and injections. Similarly, if they are calculated from the outputs of a linear state estimator, then their accuracy will deteriorate due to the quadratic relationship between voltage and power. This paper employs machine learning to perform fast and accurate flow and injection estimation in power systems that are sparsely observed by PMUs. We train a deep neural network (DNN) to learn the mapping function between PMU measurements and power flows/injections. The relation between power flows and injections is incorporated into the DNN by adding a linear constraint to its loss function. The results obtained using the IEEE 118-bus system indicate that the proposed approach performs more accurate flow/injection estimation in severely unobservable power systems compared to other data-driven methods.