Learning Based Capacitor Voltage Ripple Reduction of Modular Multilevel Converters under Unbalanced Grid Conditions with Different Power Factors

Songda Wang, Tomislav Dragicevic, Remus Teodorescu

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

Abstract

A fast and non-parameter-dependent grid-current-control method to ride through dangerous unbalanced gird condition is proposed in this paper. The grid-current references are calculated from an artificial intelligence (AI) surrogate model in order to keep the capacitor voltage at a safe level under a two phases short circuit to ground condition. And also, the circulating current reference are determined when the power factor is different when the grid fault is not serious. This machine learning network represents the relation between grid-current references and submodule capacitor voltages. The results show that this method prevents capacitor-overvoltage trips under completely short-circuited grid.
Original languageEnglish
Title of host publicationProceedings of 2020 IEEE 11th International Symposium on Power Electronics for Distributed Generation Systems
PublisherIEEE
Publication date2020
Pages531-535
ISBN (Print)9781728169903
DOIs
Publication statusPublished - 2020
EventIEEE 11th International Symposium on Power Electronics for Distributed Generation Systems - Virtual event, Dubrovnik, Croatia
Duration: 28 Sep 20201 Oct 2020

Conference

ConferenceIEEE 11th International Symposium on Power Electronics for Distributed Generation Systems
LocationVirtual event
CountryCroatia
CityDubrovnik
Period28/09/202001/10/2020
Series2020 Ieee 11th International Symposium on Power Electronics for Distributed Generation Systems (pedg)
ISSN2329-5767

Keywords

  • Modular Multilevel Converters
  • Submodule capacitor voltage, machine learning
  • Grid current control

Fingerprint Dive into the research topics of 'Learning Based Capacitor Voltage Ripple Reduction of Modular Multilevel Converters under Unbalanced Grid Conditions with Different Power Factors'. Together they form a unique fingerprint.

Cite this