Automated Measurement and Verification and Innovative Occupancy Detection Technologies

Abstract

In support of DOE's sensors and controls research, the goal of this project is to move toward integrated building to grid systems by building on previous work to develop and demonstrate a set of load characterization measurement and evaluation tools that are envisioned to be part of a suite of applications for transactive efficient buildings, built upon data-driven load characterization and prediction models. This will include the ability to include occupancy data in the models, plus data collection and archival methods to include different types of occupancy data with existing networks and a taxonomy for naming these data within a Volttron 1 agent platform. This research was conducted to: determine desired characteristics of, and technical feasibility of, new sensors that can inexpensively monitor the number of building occupants; explore how existing systems in buildings can be used to estimate the number of occupants as a function of time; and  use energy savings Measurement and Verification (M&V) methods to quantify changes in building energy performance, both with and without the use of occupancy data. Virtual SensingWe have identified more than a dozen potential data sources for virtual occupancy sensing in buildings, and collected sample data on eight of them from LBNL buildings. Specifically, we acquired data from LBNL's telephone system, its Wi-Fi infrastructure, and several sources from the IP network infrastructure. Each source has its own advantages, disadvantages, and peculiarities. A general feature of most sources is that data could be extracted as frequently as desired, and it is almost as easy to analyze results for many buildings as it is to do so for a single one. Since all hardware required is already present in buildings, the implementation cost is close to zero. The technology appears to be highly replicable and scalable. In the primary study buildings, occupancy patterns are readily visible in the data, particularly arrival, departure, and lunchtime. Weekends and holidays are also similarly quite obvious in the data. Measurement and Verification (M&V) AgentCurrent Practice: No occupancy data We used the M&V Agent that was developed for the Transactional Network project that contains a baseline model to a) predict load based on historic building load and weather data, and b) use the load predictions to quantify changes in energy use over time. Therefore, the Agent can be DisclaimerThis document was prepared as an account of work sponsored by the United States Government.

Inferring occupant counts from Wi-Fi data in buildings through machine learning

“…The occupant count information could be used for HVAC control, as in DCV [6] or model predictive control (MPC) [7], [8], since the equipment schedule and internal heat gains are correlated to the number of occupants. Additionally, occupant count is useful as the normalizing denominator in energy benchmarking, Measurement & Verification (M&V), and Fault Detection and Diagnosis (FDD) [11]. The identity and activity level information might be used to address the individual difference in thermal comfort preference [12] and to develop personalized thermal environment management [13].…”