Exploring Techniques for Monitoring Electric Power Consumption in Households

Recent works in ubiquitous computing have addressed analysis of electric power for energy conservation by detailing and studying consumption of electrical appliances. We contribute with an approach to develop techniques for fingerprinting and monitoring consumption of electric power in households. The approach builds on previous works and employs three phases: feature extraction of attributes such as real power and current harmonic contents, event detection and pattern recognition. A load library is foreseen that stores appliance characteristics as corpus data for training and recognition. We report early findings achieved using a high definition sensor directly applied to the loads showing promising results but also challenges in event detection (smaller state transitions, challenges in detecting and pairing switch on and off events). These studies are important in order to be able to address opportunities of identifying and monitoring directly at the appliance level or sensing the total load of the network. Future applications include monitoring and detailing of loads in a “balance sheet” and context aware service with advice tips for energy users. Keywordsubiquitous computing; load monitoring; fingerprinting; pattern recognition; energy awareness