Endorsing domestic energy saving behavior using micro-moment classification. (15th September 2019)

• Record Type: Journal Article
• Authors: Alsalemi, Abdullah; Ramadan, Mona; Bensaali, Faycal; Amira, Abbes; Sardianos, Christos; Varlamis, Iraklis; Dimitrakopoulos, George
• Is Part Of: Applied energy. Volume 250(2019)
• Page Start: 1302
• Page End: 1311
• Publication Date: 2019-09-15
• Subjects: Domestic energy usage -- Energy efficiency -- Big data -- Micro-moment -- Classification; Power (Mechanics) -- Periodicals; Energy conservation -- Periodicals; Energy conversion -- Periodicals; 621.042
• DOI: 10.1016/j.apenergy.2019.05.089
• Languages: English
• ISSNs: 0306-2619

Abstract

Highlights: Current environmental dilemmas necessitate developing proper energy efficiency tools. This article presents a classification system for domestic energy usage patterns. Micro-moments are short energy events used to profile domestic consumption of users. Classifiers using different parameters were trained and tested on simulated data. Ensemble bagged trees produced our highest classification accuracy of 88%. Abstract: With the ever-growing rise of energy consumption and its devastating financial and environmental repercussions, it is of utmost significance to moderate energy usage with proper energy efficiency tools. This is particularly applicable to domestic energy end-users, where an accurate profile is a prerequisite for motivating energy saving behavior. This article presents an innovative method for accurately understanding domestic energy usage patterns through a classification system. It capitalizes on the emerging concept of micro-moments, short energy-related events, and builds a comprehensive profile of end-user's energy activities with unprecedented accuracy. Micro-moments are classified based on a set of criteria per the given appliance. Five classifiers with different parameter settings were trained and tested on 10-fold cross-validated simulated data, with ensemble bagged trees topping with an accuracy of 88.0%. We also observed that linear classifiers lack in accuracy due to their inability to capture the dataset's specific structure and patterns. Fused with the other components of our framework, the proposed classification system is a novel contribution to domestic energy profiling in an effort to step energy efficiency up to the next level.