Electric-Vis: Integrating Wearable Sensing with Residential Energy Data for Improved Energy Forecasting

• Zahra Rezaei, Information Systems, University of Delaware
• Gopi Melam, Computer Science, University of Delaware

• Mathew Mauriello, Computer and Information Science, University of Delaware

Electric-Vis: Integrating Wearable Sensing with Residential Energy Data for Improved Energy Forecasting

Residential energy consumption is rising and is strongly influenced by everyday human behavior. However, most forecasting approaches rely on past power usage and weather data, overlooking wearable sensing as a source of behavioral information. As a result, the link between occupant behavior and household energy use remains underexplored, limiting personalized feedback and intervention strategies. To address this gap, Electric-Vis a web-based platform that integrates wearable sensing with residential energy data, turning this relationship into visible, actionable insights that help families reduce energy use. To support this research across multiple households, a multi-user system and admin interface were implemented, which allow each household to register independently, link one or more Fitbit accounts, and access only its own data, while an admin can oversee all users for research monitoring. This design reflects real household privacy needs while enabling the larger, more representative datasets needed for robust forecasting research.

Additionally, the platform collects household energy data at 15-minute intervals from Emporia along with wearable data from Fitbit, including steps, heart rate, and sleep, offering a detailed picture of occupant behavior. We built a comprehensive dataset by backfilling more than six months of historical wearable and energy data, providing the foundation for developing and evaluating machine learning models for residential energy forecasting. Using XGBoost with traditional features plus the added fitness tracker data, we are evaluating the potential improvement in forecasting accuracy beyond previously established predictors. Preliminary results show that 25% of predictive importance is attributed to wearable-derived features, suggesting meaningful significance of human activity in forecasting residential energy demand. With an optimized model and feedback from participating families, we will focus on refining the energy dashboard with visualizations and intervention strategies most likely to encourage reductions in household energy consumption.