Nontechnical Loss Detection of Electricity based on Neural Architecture Search in Distribution Power Networks

Lina Dong, Qi Li, Kejia Wu, Ke Fei, Chuan Liu, Ning Wang, Jun Yang, Yigui Li

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Electricity theft has always contributed a large portion of the nontechnical losses (NTLs) for the distribution networks, which usually causes severe concerns both on economics and safety of the power system operation. To cope with the rapid change of electricity theft methods, an auto NTLs detection system is proposed based on Neural Architecture Search (NAS) and Bayesian Optimization (BO). A case study utilizing NAS and BO has been performed on an electricity consumption dataset obtained from real customers of the State Grid Corporation of China. The auto-detection model achieved a F1 scores of 0.582 and an AUC of 0.919 which is competitive to the state-of-Art artificial neural network.

Original languageEnglish
Title of host publication2020 8th International Conference on Smart Grid and Clean Energy Technologies, ICSGCE 2020
PublisherIEEE
Pages143-148
Number of pages6
ISBN (Electronic)9781728157368
DOIs
Publication statusPublished - 4 Oct 2020
Event8th International Conference on Smart Grid and Clean Energy Technologies, ICSGCE 2020 - Virtual, Kuching, Malaysia
Duration: 4 Oct 20207 Oct 2020

Publication series

Name2020 8th International Conference on Smart Grid and Clean Energy Technologies, ICSGCE 2020

Conference

Conference8th International Conference on Smart Grid and Clean Energy Technologies, ICSGCE 2020
Country/TerritoryMalaysia
CityVirtual, Kuching
Period4/10/207/10/20

Keywords

  • bayesian optimization
  • electricity theft
  • neural architecture search
  • non-Technical loss

Fingerprint

Dive into the research topics of 'Nontechnical Loss Detection of Electricity based on Neural Architecture Search in Distribution Power Networks'. Together they form a unique fingerprint.

Cite this