“Multilevel Data-Driven Framework for Operating HVAC Systems to Optimize Energy and Comfort in Modern Buildings”
Architectural Engineering PhD Defense by Mehari K. Tesfay
8:00 am
Peter Kiewit Institute Room: 250
Contact:
Kelly Johnson, (402) 554-5935, kelly.johnson@unl.edu
Under the supervision of Dr. Fadi Alsaleem
Abstract: Humans spend more than 90% of their day inside buildings where their health and productivity are demonstrably linked to thermal comfort. Building comfort systems such as heating, ventilation, and air conditioning (HVAC) systems account for the largest share of U.S energy consumption. However, due to building design complexity and the variety of building occupant needs, addressing thermal comfort in buildings while reducing the high-energy cost of HVAC systems remains a difficult problem. To overcome this challenge, this dissertation presents a novel two-level data-driven framework to efficiently model and control comfort in buildings.
The first (low) level of this framework implements a control to improve the energy consumption of an HVAC system. In this implementation, an adaptive model predictive control (MPC) mechanism is developed to continuously tune the gains of the HVAC system actuator (electronic expansion valves) controller to maintain a particular parameter (superheat at the outlet of its evaporator) within a desired limit. Moreover, an adaptive setpoint hunting algorithm is implemented to select the right superheat set-point to achieve system stability. The controller performance was experimentally validated using two different HVAC systems. However, a limitation of this approach is the requirement of expensive and hard-to-install-and-maintain pressure sensors to measure superheat. To overcome this challenge, we present a novel big-data-driven approach to estimate superheat from simple temperature measurements. The approach was validated experimentally using another two different HVAC systems monitored over an extended period. The model showed good accuracy in predicting system superheat for both systems.
The second (high) level of this framework deals with modeling thermal comfort using wearable-device data to develop an intelligent controller to achieve maximum comfort. In this phase, various supervised machine-learning algorithms were evaluated to produce accurate personal thermal comfort models for each building occupant. The developed comfort models were then used to simulate an intelligent comfort controller that uses the particle swarm optimization (PSO) method to search for optimal HVAC thermostat set-point values to achieve maximum comfort. Simulation results for using the PSO algorithm were presented and showed superior performance compared to using an average thermostat set-point.
Abstract: Humans spend more than 90% of their day inside buildings where their health and productivity are demonstrably linked to thermal comfort. Building comfort systems such as heating, ventilation, and air conditioning (HVAC) systems account for the largest share of U.S energy consumption. However, due to building design complexity and the variety of building occupant needs, addressing thermal comfort in buildings while reducing the high-energy cost of HVAC systems remains a difficult problem. To overcome this challenge, this dissertation presents a novel two-level data-driven framework to efficiently model and control comfort in buildings.
The first (low) level of this framework implements a control to improve the energy consumption of an HVAC system. In this implementation, an adaptive model predictive control (MPC) mechanism is developed to continuously tune the gains of the HVAC system actuator (electronic expansion valves) controller to maintain a particular parameter (superheat at the outlet of its evaporator) within a desired limit. Moreover, an adaptive setpoint hunting algorithm is implemented to select the right superheat set-point to achieve system stability. The controller performance was experimentally validated using two different HVAC systems. However, a limitation of this approach is the requirement of expensive and hard-to-install-and-maintain pressure sensors to measure superheat. To overcome this challenge, we present a novel big-data-driven approach to estimate superheat from simple temperature measurements. The approach was validated experimentally using another two different HVAC systems monitored over an extended period. The model showed good accuracy in predicting system superheat for both systems.
The second (high) level of this framework deals with modeling thermal comfort using wearable-device data to develop an intelligent controller to achieve maximum comfort. In this phase, various supervised machine-learning algorithms were evaluated to produce accurate personal thermal comfort models for each building occupant. The developed comfort models were then used to simulate an intelligent comfort controller that uses the particle swarm optimization (PSO) method to search for optimal HVAC thermostat set-point values to achieve maximum comfort. Simulation results for using the PSO algorithm were presented and showed superior performance compared to using an average thermostat set-point.