RT journal article
T1 Thresholding methods in non-intrusive load monitoring
A1 Precioso Garcelán, Daniel
A1 Gómez-Ullate Oteiza, David
A2 Ingeniería Informática
K1 Non-intrusive load monitoring (NILM)
K1 Recurrent neural networks
K1 Convolutional neural networks
K1 Binary cross-entropy loss
K1 mean squared error loss
AB Non-intrusive load monitoring (NILM) is the problem of predicting the status orconsumption of individual domestic appliances only from the knowledge of theaggregated power load. NILM is often formulated as a classifcation (ON/OFF)problem for each device. However, the training datasets gathered by smart metersdo not contain these labels, but only the electric consumption at every time interval.This paper addresses a fundamental methodological problem in how a NILM problem is posed, namely how the diferent possible thresholding methods lead to diferent classifcation problems. Standard datasets and NILM deep learning models areused to illustrate how the choice of thresholding method afects the output results.Some criteria that should be considered for the choice of such methods are also proposed. Finally, we propose a slight modifcation to current deep learning modelsfor multi-tasking, i.e. tackling the classifcation and regression problems simultaneously. Transfer learning between both problems might improve performance on eachof them.
PB Springer
SN 0920-8542
YR 2023
FD 2023-04
LK http://hdl.handle.net/10498/28902
UL http://hdl.handle.net/10498/28902
LA eng
NO Funding for open access publishing: Universidad de Cádiz/CBUA. This research has been financed in part by the Spanish Agencia Estatal de Investigación under grants PID2021-122154NB-I00 and TED2021-129455B-I00, and by a 2021 BBVA Foundation project for research in Mathematics. He also acknowledges support from the EU under the 2014-2020 ERDF Operational Programme and the Department of Economy, Knowledge, Business and University of the Regional Government of Andalusia (FEDER-UCA18-108393).
DS Repositorio Institucional de la Universidad de Cádiz
RD 10-may-2026