Wi-Fi Based Occupancy Estimation for Optimized Operation of HVAC systems

Abstract

Occupancy information of a building can be utilized to optimize the operation of HVAC systems and to allocate resources in case of emergency. The occupancy of a building can be estimated using dedicated occupancy detection sensors like CO2 sensors, cameras, and existing data from WiFi access points. The primary shortcoming of the sensor-based approach is that additional infrastructure would have to be built to support the sensors' power delivery and data connection. Alternatively, with Wifi AP detection, no additional infrastructure would be needed as the building would already have Wifi APs installed, therefore lowering the initial cost of these detection systems. Furthermore, most sensors do not have as wide a range of coverage as Wifi APs, and multiple sensors might be needed to cover the same region as one Wifi AP. Moreover, there has been a greater concern with privacy in recent years; therefore, analyzing occupancy using camera systems might cause concerns with some individuals. In this digital age, almost everyone carries a phone or computer on their person, and their presence can be detected by the Wifi APs without any user interaction. And because buildings are already outfitted with Wifi APs, extra equipment and setup aren't required to track occupancy, and analysis can be run with a lower initial cost. For these above reasons, integration of Wifi AP-based occupancy detection would be the most effective and seamless method. We use delayed Wifi logs from the APs, including device identifiers, signal strength, and various other pieces of data. From these logs, data is picked out to be used to analyze the occupancy of a given location in the building. Pre-trained machine learning models are fed data from these logs, and occupancy prediction decisions are made, which can be utilized in optimizing the operation of HVAC systems.