A High-Fidelity Energy Monitoring and Feedback Architecture for Reducing Electrical Consumption in Buildings

Existing solutions in commercial building energy monitoring are insufficient in identifying waste or guide improvement because they only provide usage statistics in aggregate, both spatially and temporally. To significantly and sustainably reduce energy usage in buildings, we need an architecture and a system implementation that provide high-fidelity real-time visibility into each component of the building. We propose a three-tiered architecture consisting of sensing, data delivery and representation, and applications and services. We show that this layering allows us to cleanly abstract the low-level details of the myriads of disparate monitoring instruments and protocols, provide an uniform data representation interface, and enable innovation in portable building applications. We explore each layer in detail and present design decisions and findings. Building on top of this architecture, we propose an application process flow for energy data analysis and visualization, instantiated by a real deployment. This process consists of three parts – we begin by understanding and instrumenting the load tree, followed by data analysis, modeling, and disaggregation of energy usage statistics; and finally, combined with meta-data, we re-aggregate individual load usages into actionable representations for visualization and feedback to the occupants. Finally, we evaluate our architecture and process flow with a diverse class of building applications, visualizations, and deployments.

Speaker: Xiaofan Jiang, Department of Computer Science (disertation talk)

Room: RAD Lab Lounge