Full-Scale Distribution System Topology Identification Using Markov Random Field

Jian Zhao, Liang Li, Zhao Xu, Xiaoyu Wang, Haobo Wang, Xianjun Shao

Research output: Journal article publicationJournal articleAcademic researchpeer-review

53 Citations (Scopus)

Abstract

The identification of the distribution system topology is the key concern in distribution system state estimation and the precondition for its energy management. However, lacking sufficient measurement devices, full-scale identification of entire distribution grid can hardly be achievable in practice. The frequent topology changes in distribution systems impose challenges for topology identification. This paper proposes a novel topology identification method by deeply mining the data obtained from gird terminals and smart meters at end-users premises. The proposed method starts with data processing, followed by nodal correlation analysis and topology modeling based on the Markov Random Field (MRF) method, where the pseudo-likelihood method and L2 regularization theory are introduced to improve the computation efficiency while preventing the over-fitting problem. Then the iterative screening method is developed to generate the distribution system topology of medium/low-voltage distribution systems. Finally, the incremental learning and parallel programming models are proposed to implement the algorithms on single/multi-terminal. The effectiveness of the proposed model is validated on IEEE 33-node, IEEE 123-node and actual distribution systems.

Original languageEnglish
Article number9094730
Pages (from-to)4714-4726
Number of pages13
JournalIEEE Transactions on Smart Grid
Volume11
Issue number6
DOIs
Publication statusPublished - Nov 2020

Keywords

  • Distribution system
  • Markov random field
  • probabilistic graphical model
  • pseudo-likelihood
  • regularization
  • topology identification

ASJC Scopus subject areas

  • General Computer Science

Fingerprint

Dive into the research topics of 'Full-Scale Distribution System Topology Identification Using Markov Random Field'. Together they form a unique fingerprint.

Cite this