Robust event-based non-intrusive appliance recognition using multi-scale wavelet packet tree and ensemble bagging tree

Yassine Himeur,Abdullah Alsalemi,Faycal Bensaali,Abbes Amira

doi:10.1016/j.apenergy.2020.114877

Yassine Himeur, Abdullah Alsalemi + Show 2 more

Open Access

https://doi.org/10.1016/j.apenergy.2020.114877

Copy DOI

Journal: Applied Energy	Publication Date: Apr 20, 2020
Citations: 88	License type: cc-by

Affiliation: Qatar University, De Montfort University

Abstract

Providing the user with appliance-level consumption data is the core of each energy efficiency system. To that end, non-intrusive load monitoring is employed for extracting appliance specific consumption data at a low cost without the need of installing separate submeters for each electrical device. In this context, we propose in this paper a novel non-intrusive appliance recognition system based on (i) detecting events in the aggregated power signal using a novel and powerful scheme, (ii) applying multiscale wavelet packet tree to collect comprehensive energy consumption features, and (iii) adopting an ensemble bagging tree classifier along with comparing its performance with various machine learning schemes. Moreover, to validate the proposed model, an empirical investigation is conducted on two real and public energy consumption datasets, namely, the GREEND and REDD, in which consumption readings are collected at low-frequencies. In addition, a comprehensive review of recent non-intrusive load monitoring approaches has been conducted and presented, in which their characteristics, performances and limitations are described. The proposed non-intrusive load monitoring system shows a high appliance recognition performance in terms of the accuracy, F1 score and low time complexity when it has been applied to different households from the GREEND and REDD repositories, in which every house includes various domestic appliances. Obtained results have described, e.g., that average accuracies of 97.01% and 96.36% have been reached on the GREEND and REDD datasets, respectively, which outperformed almost existing solutions considered in this framework.

Highlights

Energy efficiency is considered as a demanding research topic and more attention is paid to it recently due to the benefits that can bring to the environment and society [1]
In this paper, we propose a novel non-intrusive load monitoring (NILM) architecture based on the following contributions: (1) a powerful event detection scheme is introduced, which works in the frequency-domain and deploys a filtering process in the Cesptrum space to reduce the noise generated by the electrical devices, resulting in a better detection of transitional changes occurred in the power signals; (2) an efficient feature extraction approach is introduced that is based on the multi-scale wavelet packet tree (MSWPT)
We have used the normalized cross-correlation (NCC) for measuring the similarity of the raw events and MSWPT features extracted from the original events

Summary

Introduction

Energy efficiency is considered as a demanding research topic and more attention is paid to it recently due to the benefits that can bring to the environment and society [1]. Recent studies have been reported that the best method to achieve higher energy savings in households is through monitoring each home appliance separately This can be very costly, especially when using separate plug power meters for each domestic appliance [2, 3]. In this regard, the best alternative solution is to use non-intrusive load monitoring (NILM) procedures that can separate aggregated power signal of a household or other buildings into the consumption of each specific appliance independently [4]. The former, it is mainly devoted on the use of statistical models, while the latter, which is the purpose of this framework, focalizes on three main challenges aiming at [9]: (1) detecting events and transition changes, (2)

Methods

Results

Conclusion