Learning partially observed meshed distribution grids

Harish Doddi, Deepjyoti Deka, Murti Salapaka

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This article analyzes statistical learning methods to identify the topology of meshed power distribution grids under partial observability. The learning algorithms use properties of the probability distribution of nodal voltages collected at the observed nodes. Unlike prior work on learning under partial observability, this work does not presume radial structure of the grid, and furthermore does not use injection measurements at any node. To the best of our knowledge, this is the first work for topology recovery in partially observed distribution grids, that uses voltage measurements alone. The developed learning algorithms are validated with non-linear power flow samples generated by Matpower in test grids.

Original languageEnglish (US)
Title of host publication2020 International Conference on Probabilistic Methods Applied to Power Systems, PMAPS 2020 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728128221
DOIs
StatePublished - Aug 2020
Event2020 International Conference on Probabilistic Methods Applied to Power Systems, PMAPS 2020 - Liege, Belgium
Duration: Aug 18 2020Aug 21 2020

Publication series

Name2020 International Conference on Probabilistic Methods Applied to Power Systems, PMAPS 2020 - Proceedings

Conference

Conference2020 International Conference on Probabilistic Methods Applied to Power Systems, PMAPS 2020
CountryBelgium
CityLiege
Period8/18/208/21/20

Keywords

  • distribution grids
  • low rank
  • sparsity
  • statistical learning
  • topology

Fingerprint Dive into the research topics of 'Learning partially observed meshed distribution grids'. Together they form a unique fingerprint.

Cite this