Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
A Robust Transform-Domain Deep Convolutional Network for Voltage Dip Classification
Luleå University of Technology, Department of Engineering Sciences and Mathematics, Energy Science.ORCID iD: 0000-0001-8504-494X
Department of Electrical Engineering, Chalmers University of Technology.
Luleå University of Technology, Department of Engineering Sciences and Mathematics, Energy Science.ORCID iD: 0000-0003-4074-9529
Department of Electrical Engineering, Chalmers University of Technology.
2018 (English)In: IEEE Transactions on Power Delivery, ISSN 0885-8977, E-ISSN 1937-4208, Vol. 33, no 6, p. 2794-2802Article in journal (Refereed) Published
Abstract [en]

This paper proposes a novel method for voltage dip classification using deep convolutional neural networks. The main contributions of this paper include: (a) to propose a new effective deep convolutional neural network architecture for automatically learning voltage dip features, rather than extracting hand-crafted features; (b) to employ the deep learning in an effective two-dimensional transform domain, under space-phasor model (SPM), for efficient learning of dip features; (c) to characterize voltage dips by two-dimensional SPM-based deep learning, which leads to voltage dip features independent of the duration and sampling frequency of dip recordings; (d) to develop robust automatically-extracted features that are insensitive to training and test datasets measured from different countries/regions.

Experiments were conducted on datasets containing about 6000 measured voltage dips spread over seven classes measured from several different countries. Results have shown good performance of the proposed method: average classification rate is about 97% and false alarm rate is about 0.50%. The test results from the proposed method are compared with the results from two existing dip classification methods. The proposed method is shown to out-perform these existing methods.

Place, publisher, year, edition, pages
Institute of Electrical and Electronics Engineers (IEEE), 2018. Vol. 33, no 6, p. 2794-2802
Keywords [no]
Power quality, Voltage dip, Machine learning, Deep learning, Convolutional Neural Network
National Category
Other Electrical Engineering, Electronic Engineering, Information Engineering
Research subject
Electric Power Engineering
Identifiers
URN: urn:nbn:se:ltu:diva-70217DOI: 10.1109/TPWRD.2018.2854677ISI: 000451230500023Scopus ID: 2-s2.0-85049802440OAI: oai:DiVA.org:ltu-70217DiVA, id: diva2:1236891
Note

Validerad;2018;Nivå 2;2018-12-05 (inah)

Available from: 2018-08-06 Created: 2018-08-06 Last updated: 2023-01-20Bibliographically approved
In thesis
1. Artificial Intelligence-Based Characterization and Classification Methods for Power Quality Data Analytics
Open this publication in new window or tab >>Artificial Intelligence-Based Characterization and Classification Methods for Power Quality Data Analytics
2018 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

One of the important developments in the electric power system is the fast increasing amount of data. An example of such data is formed by the voltages and currents coming from power-quality measurements. Power quality disturbances like voltage dips, harmonics and voltage transient can have a serious negative impact on the performance of equipment exposed to such disturbances. Voltage dips, short duration reductions in voltage magnitude, are especially considered as important disturbances because they regularly lead to stoppages in industrial process installations and subsequently to high costs.

The overall aim of this dissertation is the development of automatic analysis methods and other methods for extracting information from large amounts of power-quality data. This includes, methods to detect and extract event characteristics from recorded data and classify the events, for instance, based on their origins or their impact on equipment. The classification facilitates further analysis steps including reasoning and interpretation. Once the data corresponding to each class is available, a proper characterization method can be used to create more semantic data useful for information extraction. The resulting information can be used to improve the performance of the whole system, e.g., updating grid-codes, or immunity requirements of sensitive installations or processes.

This dissertation proposes different methods to fulfil each one of the above-mentioned steps. It proposes particularly a space-phasor model (SPM) of the three phase-to-neutral voltages as basis for analytic methods. The SPM is especially suitable as it is a time-domain transform without loss of any information. Another important contribution of the work is that most of the developed methods have been applied to a large dataset of about 6000 real-world voltage dips measured in existing HV and MV power networks.

The main contributions of this dissertation are as follows:

A complete framework has been proposed for automatic voltage quality analysis based on the SPM. The SPM has been used before, but this is the first time it has been used in a framework covering a range of voltage quality disturbances. A Gaussian-based anomaly detection method is used to detect and extract voltage quality disturbances. A principal component analysis (PCA) algorithm is used for event characterization. The obtained single-event characteristics are used to extract additional information like origin, fault type and location. 

Two deep learning-based voltage dip classifier has been developed. In both classifier a 2D convolutional neural network (2D-CNN) architecture has been employed to perform automatic feature extraction task. The soft-max activation function fulfills supervised classification method in first classifier. The second classifier uses a semi-supervised classification method based on generative-discriminative model pairs in active learning context.

The same SPM was shown to enable the effective extraction of dip characteristics for multi-stage voltage dips. Applying the k-means clustering algorithm, the event is clustered into its individual stages. For each stage of the dip, a logistic regression algorithm is used to characterize that stage. The proposed method offers a new solution to the problem with transition segments that is one of the main challenges of existing methods for characterization of multi-stage dips.  

It is also shown in the dissertation that the SPM is an effective method for voltage transient analysis. It is possible to extract corresponding sample data and get appropriate single-event characteristics.

A systematic way has been developed and applied for comparing different sets of voltage dip characteristics. With this method, both measured and synthetic voltage dips are applied to generic models of sensitive loads. The best set of characteristics is the one most accurately reproducing the behavior of equipment when exposed to measured voltage dips.

The dissertation further contains a number of practical applications of the before-mentioned theoretical contributions: a proposal to an international standard-setting group; energy storage for voltage-dip ride-through of microgrids; impact of different voltage dips on wind-power installations.

Place, publisher, year, edition, pages
Luleå: Luleå University of Technology, 2018. p. 213
Series
Doctoral thesis / Luleå University of Technology 1 jan 1997 → …, ISSN 1402-1544
Keywords
Power System, power Quality, Voltage Dip, Big Data, Deep Learning, Machine Learning, Active Learning, Consensus Contriol
National Category
Other Electrical Engineering, Electronic Engineering, Information Engineering
Research subject
Electric Power Engineering
Identifiers
urn:nbn:se:ltu:diva-71413 (URN)978-91-7790-250-8 (ISBN)978-91-7790-251-5 (ISBN)
Public defence
2018-12-07, Skellefteå Hörsal A193, Skellefteå, 10:00 (English)
Opponent
Supervisors
Available from: 2018-11-01 Created: 2018-11-01 Last updated: 2019-03-04Bibliographically approved

Open Access in DiVA

fulltext(883 kB)742 downloads
File information
File name FULLTEXT01.pdfFile size 883 kBChecksum SHA-512
49b61f9a59c53e725af3104f6ec26cb5a28feaf9432160f22d84f09a4404353b05637fefc0ed1653f1aed4750d2cb6f7351494127587a96b2113772d2b7997f5
Type fulltextMimetype application/pdf

Other links

Publisher's full textScopus

Authority records

Bagheri, AzamBollen, Math

Search in DiVA

By author/editor
Bagheri, AzamBollen, Math
By organisation
Energy Science
In the same journal
IEEE Transactions on Power Delivery
Other Electrical Engineering, Electronic Engineering, Information Engineering

Search outside of DiVA

GoogleGoogle Scholar
Total: 742 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

doi
urn-nbn

Altmetric score

doi
urn-nbn
Total: 395 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf