Authors:
Roman Jonetzko
;
Matthias Detzler
;
Klaus-Uwe Gollmer
;
Achim Guldner
;
Marcel Huber
;
Rainer Michels
and
Stefan Naumann
Affiliation:
Trier University of Applied Sciences, Germany
Keyword(s):
Pattern Recognition, Non-intrusive Load Monitoring, Classification, Fourier Descriptors.
Related
Ontology
Subjects/Areas/Topics:
Energy and Economy
;
Energy Monitoring
;
Energy-Aware Systems and Technologies
;
Smart Cities
;
Smart Sensor-Based Networks and Applications
Abstract:
The number of electronic devices in households as well as in industrial workplaces is continuously growing because
of progress in automation. Identifying unusual operating behavior, detecting device failures in advance,
and recognizing energy saving potentials are key features to improve the reliability, safety, and profitability
of those systems. Facing these tasks, todays research is focused inter alia on a non-intrusive load monitoring
approach, where the electrical signal is measured at a central point with modern hardware and processed by
pattern recognition algorithms. Thus, we developed a smart meter prototype with a high sampling frequency,
which allows for continuous measurement of the current and voltage from three-phase power lines. Besides
this, in this paper we describe the usage of current-only measurement data (simple and safe installation using
current transformers) with which we were able to classify state changes of a mobile air-conditioner with the
help of Fourier
descriptors as well as with additional voltage measurement.
(More)