Research Trends and Applications of PMUs

Abstract

This work is a survey of current trends in applications of PMUs. PMUs have the potential to solve major problems in the areas of power system estimation, protection, and stability. A variety of methods are being used for these purposes, including statistical techniques, mathematical transformations, probability, and AI. The results produced by the techniques reviewed in this work are promising, but there is work to be performed in the context of implementation and standardization. As the smart grid initiative continues to advance, the number of intelligent devices monitoring the power grid continues to increase. PMUs are at the center of this initiative, and as a result, each year more PMUs are deployed across the grid. Since their introduction, myriad solutions based on PMU-technology have been suggested. The high sampling rates and synchronized measurements provided by PMUs are expected to drive significant advancements across multiple fields, such as the protection, estimation, and control of the power grid. This work offers a review of contemporary research trends and applications of PMU technology. Most solutions presented in this work were published in the last five years, and techniques showing potential for significant impact are highlighted in greater detail. Being a relatively new technology, there are several issues that must be addressed before PMU-based solutions can be successfully implemented. This survey found that key areas where improvements are needed include the establishment of PMU-observability, data processing algorithms, the handling of heterogeneous sampling rates, and the minimization of the investment in infrastructure for PMU communication. Solutions based on Bayesian estimation, as well as those having a distributed architectures, show great promise. The material presented in this document is tailored to both new researchers entering this field and experienced researchers wishing to become acquainted with emerging trends.

Distributed MPC Method for Power Systems Frequency Control: A Particle Filter Integrated Model for Uncertainty Quantification

“…PMUs have relatively high sampling rates, in some cases up to 120 frames per second (fps) [21]. These sampling rates combined with the synchronization of measurements, make it possible to derive input-output pairs that can be used to build mathematical models [22].…”

“…Sampling rate limitations are in part driven by the capacity of communi-Version April 18, 2023 submitted to Energies 6 of 21 cation networks. High PMU sampling rates could overwhelm communication networks in terms of data transfer and data storage [11,22]. With these limitations in mind, 30 fps seems to offer a balance between practicality, and estimation accuracy.…”

“…Before a follow-up load shedding stage is executed, synchronisation between the machines is considered. Surprisingly, synchronisation is often overlooked in frequency stability solutions in literature [22]. In this work, machine rotor angles are assumed to be tracked via relays at the same sampling rate used for frequency, and the difference between the rotor angles, ∆θ, together with the predicted state of frequency, are used to determine the need for follow-up load shedding stages.…”

“…With these challenges in mind, it becomes evident that a new approach to inertia estimation is needed. The sampling rates and synchronisation of PMUs make it possible to utilize system identification algorithms to estimate power system parameters [22]. One such approach, subspace estimation, has delivered encouraging results in power systems applications [23,32,33].…”

State Estimation in Partially Observable Power Systems via Graph Signal Processing Tools

“…Dynamic state estimation utilizes measurements at different time points [ 14 , 15 ], but needs fast scan rates to capture the dynamics, and is also based on restrictive stationary system assumptions [ 14 ]. PSSE that uses data from smart meters and phasor measurement units (PMUs) to overcome observability issues [ 10 , 16 , 17 , 18 ] is usually inapplicable, due to the limited deployment of these sensors [ 19 ], and their initial installment cost [ 20 , 21 ]. Sparse signal recovery methods [ 5 ] use matrix completion to estimate the states under low observability conditions.…”