Next Article in Journal
Averaged Optimization and Finite-Time Thermodynamics
Next Article in Special Issue
A New Pooling Approach Based on Zeckendorf’s Theorem for Texture Transfer Information
Previous Article in Journal
Wigner Time-Delay and Distribution for Polarization Interaction in Strongly Coupled Semiclassical Plasmas
Previous Article in Special Issue
Deep Learning for Stock Market Prediction
Article

On the Use of Concentrated Time–Frequency Representations as Input to a Deep Convolutional Neural Network: Application to Non Intrusive Load Monitoring

1
Laboratoire IBISC (Informatique, BioInformatique, Systèmes Complexes), EA 4526, University Evry/Paris-Saclay, 91020 Evry CEEEE, France
2
Institut de Recherche en Energie Electrique de Nantes Atlantique (IREENA), EA 4642, University of Nantes, 44602 Saint-Nazaire, France
*
Author to whom correspondence should be addressed.
Entropy 2020, 22(9), 911; https://doi.org/10.3390/e22090911
Received: 29 July 2020 / Revised: 14 August 2020 / Accepted: 17 August 2020 / Published: 19 August 2020
(This article belongs to the Special Issue Information Transfer in Multilayer/Deep Architectures)
Since decades past, time–frequency (TF) analysis has demonstrated its capability to efficiently handle non-stationary multi-component signals which are ubiquitous in a large number of applications. TF analysis us allows to estimate physics-related meaningful parameters (e.g., F0, group delay, etc.) and can provide sparse signal representations when a suitable tuning of the method parameters is used. On another hand, deep learning with Convolutional Neural Networks (CNN) is the current state-of-the-art approach for pattern recognition and allows us to automatically extract relevant signal features despite the fact that the trained models can suffer from a lack of interpretability. Hence, this paper proposes to combine together these two approaches to take benefit of their respective advantages and addresses non-intrusive load monitoring (NILM) which consists of identifying a home electrical appliance (HEA) from its measured energy consumption signal as a “toy” problem. This study investigates the role of the TF representation when synchrosqueezed or not, used as the input of a 2D CNN applied to a pattern recognition task. We also propose a solution for interpreting the information conveyed by the trained CNN through different neural architecture by establishing a link with our previously proposed “handcrafted” interpretable features thanks to the layer-wise relevant propagation (LRP) method. Our experiments on the publicly available PLAID dataset show excellent appliance recognition results (accuracy above 97%) using the suitable TF representation and allow an interpretation of the trained model. View Full-Text
Keywords: non-intrusive load monitoring (NILM); time–frequency representation (TFR); deep learning; convolutional neural network (CNN); synchrosqueezing; Layer wise relevance propagation (LRP) non-intrusive load monitoring (NILM); time–frequency representation (TFR); deep learning; convolutional neural network (CNN); synchrosqueezing; Layer wise relevance propagation (LRP)
Show Figures

Figure 1

MDPI and ACS Style

Houidi, S.; Fourer, D.; Auger, F. On the Use of Concentrated Time–Frequency Representations as Input to a Deep Convolutional Neural Network: Application to Non Intrusive Load Monitoring. Entropy 2020, 22, 911. https://doi.org/10.3390/e22090911

AMA Style

Houidi S, Fourer D, Auger F. On the Use of Concentrated Time–Frequency Representations as Input to a Deep Convolutional Neural Network: Application to Non Intrusive Load Monitoring. Entropy. 2020; 22(9):911. https://doi.org/10.3390/e22090911

Chicago/Turabian Style

Houidi, Sarra, Dominique Fourer, and François Auger. 2020. "On the Use of Concentrated Time–Frequency Representations as Input to a Deep Convolutional Neural Network: Application to Non Intrusive Load Monitoring" Entropy 22, no. 9: 911. https://doi.org/10.3390/e22090911

Find Other Styles
Note that from the first issue of 2016, MDPI journals use article numbers instead of page numbers. See further details here.

Article Access Map by Country/Region

1
Back to TopTop