Developing a Dynamic Control Algorithm to Improve Ventilation Efficiency in a University Conference Room; Impacts attributed to the Adversarial Manipulation of Smart HVAC Systems on Building Occupants

Jabbour, Jason, School of Engineering and Applied Science, University of Virginia
Heydarian, Arsalan, EN-Eng Sys and Environment, University of Virginia
Small, Arthur, EN-Eng Sys and Environment, University of Virginia
Earle, Joshua, EN-Engineering and Society, University of Virginia

HVAC systems are needed to ventilate buildings to maintain healthy indoor environments. Such HVAC systems foster healthy and comfortable indoor environments by maintaining a constant temperature and increasing indoor air quality (IAQ). Current HVAC systems in commercial buildings successfully achieve constant indoor temperature and healthy IAQ levels. Still, this objective is reached at the cost of high energy usage, according to the U.S. Department of Energy. With the call for the creation of more advanced HVAC systems and the increased time Americans spend indoors due to the SARS-CoV-2, this research presents a control algorithm (CA) to optimize the energy usage of HVAC systems. The proposed control algorithm recommends a ventilation state (on/off) every 15 minutes by modeling future IAQ values and learning the occupancy pattern within a ventilation zone. Such a control algorithm improves the current HVAC frameworks because of the energy saved by not ventilating an enclosed space during the weekends or early/late hours of the day when that space is unoccupied. This research was conducted within a 490 ft2 conference room in the University of Virginia’s Link Lab. Throughout this research, IAQ is defined by the levels of the three primary toxins found within enclosed spaces: Carbon dioxide (CO2), particulate matter (PM2.5), and volatile organic compounds (VOCs). To collect such metrics, the Link Lab was equipped with Awair brand sensors which collected temperature, CO2, PM2.5, and VOC readings over two months (October to December 2021). The control algorithm converts IAQ to a dollar productivity cost. Particularly, a regression equation for each IAQ metric is used to calculate a productivity cost. The control algorithm also utilizes the energy cost spent by the HVAC system. The energy cost is calculated based on fan energy, heating/cooling energy, dehumidifying energy, and zone reheat energy. Finally, the control algorithm is fed occupancy data about the given space. Since the testbed involves a conference room within a university setting, building occupants are assumed to maintain a relatively consistent routine during a semester. A motion sensor was used to collect data from within the conference room over the same two-month period to learn the patterns of the building occupants. The collected data was then passed into a machine learning model to learn the schedule of building occupants visiting the conference room. By utilizing the day-of-week, hour-of-week, and quarter-hour as features, the model learned that the conference room was generally vacant during weekends and early/late hours of the day. By incorporating IAQ productivity cost, HVAC energy cost, and the occupancy status of the conference room, an objective function was constructed to maximize IAQ and minimize energy cost. The STS portion of this paper presents the risks that could surface due to the creation of “smart” HVAC systems. The progression of physical systems gaining a connection to the internet creates unique opportunities for adversaries to exploit vulnerabilities within the system and potentially harm building occupants. This paper will first describe IoT devices and the methods by which they could be manipulated. Next, this paper will explore the algorithmic component of a “smart” HVAC system. With each HVAC system programmed to optimize specific metrics, the paper will explore how some adversarial groups could leverage mathematics to achieve unethical behavior. Building off the dangers of an HVAC system as an IoT device and the malicious intent of individuals to alter the behavior of a “smart” HVAC system, this paper will present a real-world example of a parallel system. Introducing an IoT device into large-scale systems such as HVAC systems elevates the risk usually associated with IoT devices. Evolving HVAC systems into “smart” HVAC systems changes the dynamic and precautions needed to defend such systems since the physical safety of building occupants is put in dispute. Finally, this paper will explore the types of individuals or groups that could be motivated to manipulate or attack a “smart” HVAC system. Through these components, thoughtful caution will be set about the utilization of HVAC systems equipped with mathematical models and technological advancements within buildings. This paper presents a cautionary warning from several perspectives to highlight the need for deeper studies to be conducted among a collaboration of fields before such a new technological change could be comfortably and widely accepted among commercial buildings.

BS (Bachelor of Science)
HVAC, Smart HVAC System, IAQ, Indoor Air Quality, Adversarial Manipulation of Smart HVAC, Ventilation Efficiency, Conference Room IAQ

School of Engineering and Applied Science
Bachelor of Science in Systems Engineering
Technical Advisor: Arsalan Heydarian, Arthur Small
STS Advisor: Joshua Earle
Technical Team Members: Matthew Caruso, Caleb Neale, Alden Summerville, Avery Walters

Issued Date: