Buildings account for a large portion of the total energy consumption in Canada. Most of this energy is consumed to maintain occupants' thermal and visual comfort. Thus, understanding the building occupancy is critical for designing energy-efficient building controls, maximizing human comfort, and creating a safe and healthy workplace during the pandemic. Despite the importance of incorporating occupancy information in the control of HVAC, lights, and window blinds, this information is often unavailable or inaccurate. Machine learning models can help fuse data from various sources (e.g., WiFi association logs, prox cards, cameras) to estimate occupancy at different spatial and temporal resolutions. Leveraging reinforcement learning algorithms, it is possible to determine optimal (energy-efficient) control policies which satisfy stringent safety requirements.

This project aims to analyze the estimated building occupancy levels, whole-building energy consumption patterns, and operational data from the Building Management System (BMS) to 

1. Refine the estimated number of occupants in each building and identify/address data quality issues. This is an important task as there is not a one-to-one mapping between WiFi users and building occupants. The refined occupancy information will be useful for monitoring building occupancy and issuing warnings if social distancing rules are violated (during the pandemic).
   
   
2. Understand what percentage of the whole-facility energy use can be attributed to occupants, and utilizing this information to optimize the operation of HVAC and lighting systems. The optimal control policy is expected to reduce the building energy use while satisfying comfort requirements.