Machine Learning-Enabled Distribution Network Phase Identification

Hosseini, Zohreh S.; Khodaei, Amin; Paaso, Aleksi

doi:10.1109/tpwrs.2020.3011133

Cited by 55 publications

(36 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• Hosseini et al [3] use a Finite Impulse Response (FIR) high-pass filter to filter out the low frequencies in each time series. In this work we choose to use the time difference transformation as it is a non-parametric technique that is computationally fast (runs in linear time) and acts as a high-pass filter.…”

Section: A Stationarity: Second Order Time Differencementioning

confidence: 99%

“…Knowing the phase connectivity of loads is important for keeping a distribution system operating under balanced conditions where the amount of power flowing on each phase is as close to equal as possible [1]. An unbalanced system can result in issues such as lower operating efficiency, component degradation (such as that of secondary distribution transformers), and overheating [2] [3] to name a few. However, maintaining accurate records of phase connections can be made difficult due to events such as maintenance, system reconfigurations, emergency power restorations (such as due to extreme weather events), and potential cyberattacks [4] [5].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Denoising with Singular Value Decomposition for Phase Identification in Power Distribution Systems

Zaragoza¹,

Rao²

2022

Preprint

View full text Add to dashboard Cite

Phase identification is the problem of determining what phase(s) that a load is connected to in a power distribution<br>system. However, real world sensor measurements used for phase identification have some level of noise that can hamper the ability to identify phase connections using data driven methods. Knowing the phase connections is important to keep the distribution system balanced so that parts of the system aren’t overloaded which can lead to inefficient operations, accelerated component degradation, and system destruction at worst. We use Singular Value Decomposition (SVD) with the optimal Singular Value Hard Threshold (SVHT) as part of a feature engineering pipeline to denoise data matrices of voltage magnitude measurements. This approach results in a reduction in frobenius error and an increase in average phase identification accuracy over a year of time series data. K-medoids clustering is used on the denoised voltage magnitude measurements to perform phase identification.<br>

show abstract

Section: A Stationarity: Second Order Time Differencementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Denoising with Singular Value Decomposition for Phase Identification in Power Distribution Systems

Zaragoza¹,

Rao²

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…• Hosseini et al [3] use a Finite Impulse Response (FIR) high-pass filter to filter out the low frequencies in each time series. In this work, we choose to use the time difference transformation, as it is a non-parametric technique that is computationally fast (runs in linear time) and acts as a high-pass filter [18].…”

Section: A Stationarity: Second Order Time Differencementioning

confidence: 99%

Denoising with Singular Value Decomposition for Phase Identification in Power Distribution Systems

Zaragoza¹,

Rao²

2021

Preprint

View full text Add to dashboard Cite

Phase identification is the problem of determining what phase(s) that a load is connected to in a power distribution system. However, real-world sensor measurements used for phase identification have some level of noise that can hamper the ability to identify phase connections using data-driven methods. Knowing the phase connections is important to keep the distribution system balanced so that parts of the system are not overloaded, which can lead to inefficient operations, accelerated component degradation, and system destruction at worst. We use Singular Value Decomposition (SVD) with the optimal Singular Value Hard Threshold (SVHT) as part of a feature engineering pipeline to denoise data matrices of voltage magnitude measurements. This approach reduces Frobenius error and increases the average phase identification accuracy over a year of time series data. K- medoids clustering is used on the denoised voltage magnitude measurements to perform phase identification.<br><br>

show abstract

“…In [7], a careful and extensive review of cyber-physical attacks and security issues in future power grids is presented, categorizing existing cyber-physical attacks and defense approaches, as well as future challenges and opportunities. In [8], the contribution of machine learning technologies to future power grids is presented, aiming at their stability and management, and considering the profile of residential consumers and the dynamic behavior of the power grid. In [9], an exhaustive review on the application of blockchain to ensure cybersecurity in power grids is presented, introducing the latest insights, as well as implementation architectures and techniques.…”

Section: Introductionmentioning

confidence: 99%

Review of a Disruptive Vision of Future Power Grids: A New Path Based on Hybrid AC/DC Grids and Solid-State Transformers

et al. 2021

View full text Add to dashboard Cite

Power grids are evolving with the aim to guarantee sustainability and higher levels of power quality for universal access to electricity. More specifically, over the last two decades, power grids have been targeted for significant changes, including migration from centralized to decentralized paradigms as a corollary of intensive integration of novel electrical technologies and the availability of derived equipment. This paper addresses a review of a disruptive vision of future power grids, mainly focusing on the use of hybrid AC/DC grids and solid‑state transformers technologies. Regarding hybrid AC/DC grids in particular, they are analyzed in detail in the context of unipolar and bipolar DC grids (i.e., two‑wire or three‑wire DC grids), as well as the different structures concerning coupled and decoupled AC configurations with low‑frequency or high‑frequency isolation. The contextualization of the possible configurations of solid‑state transformers and the different configurations of hybrid transformers (in the perspective of offering benefits for increasing power quality in terms of currents or voltages) is also analyzed within the perspective of the smart transformers. Additionally, the paper also presents unified multi‑port systems used to interface various technologies with hybrid AC/DC grids, which are also foreseen to play an important role in future power grids (e.g., the unified interface of renewable energy sources and energy storage systems), including an analysis concerning unified multi‑port systems for AC or DC grids. Throughout the paper, these topics are presented and discussed in the context of future power grids. An exhaustive description of these technologies is made, covering the most relevant and recent structures and features that can be developed, as well as the challenges for the future power grids. Several scenarios are presented, encompassing the mentioned technologies, and unveiling a progressive evolution that culminates in the cooperative scope of such technologies for a disruptive vision of future power grids.

show abstract

Machine Learning-Enabled Distribution Network Phase Identification

Cited by 55 publications

References 22 publications

Denoising with Singular Value Decomposition for Phase Identification in Power Distribution Systems

Denoising with Singular Value Decomposition for Phase Identification in Power Distribution Systems

Denoising with Singular Value Decomposition for Phase Identification in Power Distribution Systems

Review of a Disruptive Vision of Future Power Grids: A New Path Based on Hybrid AC/DC Grids and Solid-State Transformers

Contact Info

Product

Resources

About