Nonintrusive load monitoring based on recurrence plots and deep learning in the context of smart homes

• Grant number: 19/15192-9
• Support Opportunities: Regular Research Grants
• Start date: November 01, 2019
• End date: January 31, 2022
• Field of knowledge: Engineering - Electrical Engineering - Electrical, Magnetic and Electronic Measurements, Instrumentation
• Principal Investigator: Ricardo Augusto Souza Fernandes
• Grantee: Ricardo Augusto Souza Fernandes
• Host Institution: Centro de Ciências Exatas e de Tecnologia (CCET). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil
• Associated researchers: Denis Vinicius Coury ; Tatiane Cristina da Costa Fernandes

Abstract

Currently, there is great concern with energy management, especially in large consumer centers. In this sense, many researches have been aimed at energy management for residential consumers, since these are aimed at obtaining solutions that allow nonintrusive load monitoring and identification, as well as better efficiency in the consumption of electricity. According to this context, this research project seeks to improve the nonintrusive load monitoring for residential consumers, considering that changes in load state occur over time. Therefore, it is possible to highlight one of the great challenges found in the literature, which is the identification of non-linear patterns in the classification of time series, commonly present in dynamic systems or stochastic processes. Frequently, these phenomena are recurrent, so that certain regions of their state space are often visited. Moreover, the visualization of these behaviors is usually very difficult in the time domain. Following this premise, the present project aims to conduct an extensive study on the joint use of Recurrence Plots and Deep Learning techniques for the classification of time series. This way, modifications will be proposed in the Recurrence Plots algorithm so that it becomes suitable for the purpose of nonintrusive load monitoring. This stage of pre-processing, performed through the Recurrence Plots, will be analyzed considering data obtained at different sampling rates. Thus, a dataset obtained through experiments of the Laboratory of Applied Artificial Intelligence (LIAA) of the Federal University of São Carlos will be considered. In the sequence, Deep Learning techniques will be considered for the detection of events (changes of state) and classification of the loads. (AU)

Scientific publications

(4)

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

MORAES, ANDERSON L.; FERNANDES, RICARDO A. S.; IEEE. Recurrence Plots: A Novel Feature Engineering Technique to Analyze Power Quality Disturbances. 2020 IEEE POWER & ENERGY SOCIETY GENERAL MEETING (PESGM), v. N/A, p. 5-pg., 2020-01-01. (19/15192-9)

CAVALCA, DIEGO L.; FERNANDES, RICARDO A. S.; IEEE. Recurrence Plots and Convolutional Neural Networks Applied to Nonintrusive Load Monitoring. 2020 IEEE POWER & ENERGY SOCIETY GENERAL MEETING (PESGM), v. N/A, p. 5-pg., 2020-01-01. (19/15192-9)

CAVALCA, DIEGO L.; FERNANDES, RICARDO A. S.. Deep Transfer Learning-Based Feature Extraction: An Approach to Improve Nonintrusive Load Monitoring. IEEE ACCESS, v. 9, p. 139328-139335, 2021. (19/15192-9)

DE MORAES, ANDERSON LUIS; COURY, DENIS VINICIUS; FERNANDES, RICARDO AUGUSTO SOUZA. Power Quality Disturbances Segmentation: An Approach Based on Gramian Angular Field. ELECTRIC POWER COMPONENTS AND SYSTEMS, v. N/A, p. 16-pg., 2023-09-12. (21/04872-9, 23/00182-3, 19/15192-9)