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
Methods to identify broken neutral fault in LV distribution grids by using existing smart meters infrastructure
KTH, School of Industrial Engineering and Management (ITM).
2019 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

The vast majority of Swedish utility network customers for nearly a decade has been supported by Advanced Metering Management (AMM) systems, including smart meters. Vattenfall´s modern smart meters enable a new level of monitoring for LV networks and improved MV network supervision. Therefore, improving power quality, fault detection and outage management functionalities are just some of the areas that smart metering systems can contribute to. One of the typical faults in LV networks is when the neutral conductor is broken or loose at either the network or the load side of the meter. The situation of lost continuity of the neutral conductor may damage the connected load or create hazardous touch voltages at equipment body. Since there is a big potential to supervise LV networks with assistance of the end-customer smart meters, Vattenfall wants to take further advantage of such data. The value is to bring in event information from the smart meters in order to contribute to a better and more efficient monitoring of the LV and MV network. The goal of this project is to analyze the behavior of the LV grid under broken neutral fault conditions and propose effective methods (algorithms) to identify loose neutral situation based on end-customer meter readings (disturbance events).

Based on previous literature review and studies conducted for broken neutral fault detection, phase to neutral voltages has been proved that can be a useful indicator to detect the fault, since there is a clear pattern during the fault. However, the voltages-based method is not always effective, such as during periods when the load is almost balanced among the three phases or when the load magnitude is not high enough. This is the reason why other electrical parameters could be useful as well to detect the fault, except from the phase to neutral voltages. This study adds a great value into the study of broken neutral fault in low voltage grids since no previous work has been found where dynamic load profiles are modelled and simulated.

The broken neutral fault study has started with the creation of dynamic load profiles that has been used in MATLAB/SIMULINK to model inductive linear load with or without the integration of single-phase PV assets. Furthermore, non-linear load has been investigated during BN fault in a single-customer model, where three case studies with different percentages of nonlinear load integration into the system have been included. Later, a 7-customer low voltage rural grid has been modelled where not only broken neutral but also phase loss and short circuit faults have been modeled and simulated. 9 different locations for Broken Neutral and Phase Loss faults and 7 locations for Short Circuit fault, 4 seasons with different load profiles and 4 different PV integration combinations with single and three-phase assets were considered. It has been proved that the combination of different electrical parameters and not only phase to neutral voltages can improve significantly the detection of broken neutral fault, not only on the DSO side but also at the customer side, with the use of smart meter data.

Last but not least, part of this study has been to use the data that have been produced from the simulations to train a machine learning model that can accurately detect broken neutral fault. For that reason, a Proof of Concept using different machine learning classification methods as well as neural networks have been trained and tested, based on large amount of data, has been proposed. Bagged decision trees have been found as the most accurate method.

It is important to highlight that due to data confidentiality issues, specific values and thresholds that have been set in the algorithms that are currently used or proposed cannot be published.

Place, publisher, year, edition, pages
2019. , p. 108
Series
TRITA-ITM-EX ; 2019:572
National Category
Engineering and Technology
Identifiers
URN: urn:nbn:se:kth:diva-264280OAI: oai:DiVA.org:kth-264280DiVA, id: diva2:1372876
External cooperation
Vattenfall AB
Supervisors
Examiners
Available from: 2019-11-25 Created: 2019-11-25 Last updated: 2019-11-25Bibliographically approved

Open Access in DiVA

fulltext(10351 kB)5 downloads
File information
File name FULLTEXT01.pdfFile size 10351 kBChecksum SHA-512
2eec98077223f1469ecaa2071fafc34fe0f479c921f2ee32ca4bf9f98ded855f4d3e7254e41fabe28068dcf375e0417e280f0dbbcf4ed0c3ba4ef5f35d2117a7
Type fulltextMimetype application/pdf

By organisation
School of Industrial Engineering and Management (ITM)
Engineering and Technology

Search outside of DiVA

GoogleGoogle Scholar
Total: 5 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

urn-nbn

Altmetric score

urn-nbn
Total: 13 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