Electrical Appliance Classification using Deep Convolutional Neural Networks on High Frequency Current Measurements

Daniel Jorde, Thomas Kriechbaumer, Hans Arno Jacobsen

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

6 Scopus citations

Abstract

Monitoring the energy demand of appliances can raise consumer awareness and therefore reduce energy consumption. Using a single-point measurement of mains energy consumption can keep costs and hardware complexity to a minimum. This data stream of raw voltage and current measurements can be used in machine learning tasks to extract information. We apply Deep Convolutional Neural Networks on an electrical appliance classification task, using raw high frequency start up events from two datasets. We further present Data Augmentation techniques to improve the model performance and evaluate different data normalization techniques. We achieve a perfect classification on WHITED and a Fl-Score of 0.69 on PLAID.

Original languageEnglish
Title of host publication2018 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781538679548
DOIs
StatePublished - 24 Dec 2018
Event2018 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2018 - Aalborg, Denmark
Duration: 29 Oct 201831 Oct 2018

Publication series

Name2018 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2018

Conference

Conference2018 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2018
Country/TerritoryDenmark
CityAalborg
Period29/10/1831/10/18

Fingerprint

Dive into the research topics of 'Electrical Appliance Classification using Deep Convolutional Neural Networks on High Frequency Current Measurements'. Together they form a unique fingerprint.

Cite this