Simulating Appliance-Based Power Consumption Records for Energy Efficiency Awareness

This is the authors' abstract. We don't add key points for this paper.

The collection and analysis of big data has become a ubiquitous process in numerous fields, including energy efficiency. The combination of smart meters that provide high-frequency readings of consumption and enhanced data analytics have enabled studies to uncover energy usage patterns of households. However, an accurate understanding of consumer behavior requires a more comprehensive approach. Henceforth, we present a novel data simulator that generates appliance-based datasets based on real data. By proper fusion of real smart meter data and periodic energy consumption habits, which are represented by micro-moments, we were able to simulate realistic domestic energy consumption scenarios. Pre-processing the aggregated readings and the application of a k-means clustering algorithm, user actions were detected (switch on/off) per appliance. Based on these records and using a apriori itemset extraction algorithm we obtained appliance usage patterns. A minute-averaged dataset was simulated for three rooms and nine appliances, monitored for a 47-month period. We managed to extract frequent usage patterns at appliance-level, taking into account contextual parameters such as room occupancy. Moreover, the simulated datasets can be expanded to any size to a variety of energy saving applications such as micro-moment classification and goal-based recommender systems for energy saving.