Control of personalized thermal conditioning systems

Metzmacher, Henning; van Treeck, Christoph Alban (Thesis advisor); Rumpe, Bernhard (Thesis advisor)

Aachen : RWTH University Aachen (2020, 2021)
Dissertation / PhD Thesis

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2020

Abstract

Global warming is a major aspect of climate change and is caused primarily by the emission of greenhouse gases through industrial processes and the burning of fossil fuels. The energy sector comprises 72% of global greenhouse gas emission which can be divided into electricity and heating demand (31.0%), transportation (15.0%), manufacturing and construction (12.4 %), fuel and combustion (8.4%) and fugitive emissions (5.2%) [Pac+14]. A large part of energy consumption of residential and commercial buildings as well as the transportation sector can be attributed to modern air conditioning [POP08; FR00; Lu+05]. Perez et al. state that heating, ventilation and air conditioning (HVAC) systems are the most energy consuming devices of final energy use in developed countries [Pér+11]. Reduction in energy consumption for air conditioning is therfore an important contributor to the mitigation of global warming.Currently, air conditioning is commonly done by heating or cooling the entire air volume of a space. While in modern consumer vehicles there are settings for zonal heating or cooling, control heuristics generally run at a global or semi-global level. Similarly, air conditioning in buildings is mostly done by regarding rooms as single zones. Conditioning an entire space, however, is not always the most efficient way to maintain a desirable temperature for occupants or passengers. This is because energy required to change the temperature of air that is not in direct contact with the human body is effectively wasted. A much more efficient way to maintain a desirable thermal state of a person is to directly heat or cool segments of the body without changing the overall air temperature. This method, which is referred to as "personalized thermal conditioning" utilizes user specific information acquired through sensor input, direct user feedback and thermophysiological simulation.This thesis focuses on the development of a personalized thermal conditioning system that controls local heating and cooling actuators in close proximity to the human body. The system uses segment-wise skin temperature information in combination with internal models in order to estimate the current state of thermal comfort. Skin temperatures are acquired using a thermal infrared camera and a face and pose tracking algorithm. Emperical models are trained on-the-fly using direct user feedback which is correlated with corresponding skin temperature measurements. Specifically, models used in this work are an existing comfort model proposed by Zhang [Zha+10b] which is kept fixed and serves as a reference model, an adaptive version of the Zhang model where coefficients are modified using Monte Carlo sampling as well as decision tree, support vector machine and multilayer perceptron models which are all trained bottom-up with no prior information. The system is designed in such a way that different software and hardware components communicate asynchronously over a central communication server. Here, a simple key-value storage is used to exchange data between sensor and actuator software, numerical models, visualization tools and external software. The software components connected to the data server include a vision component that combines user face and pose tracking with thermal infrared skin temperature measurements, components that serve as adapters for sensor and actuator hardware, a user component which provides a user feedback interface along with capabilities to learn thermal comfort models and co-simulation adapters used to integrate human thermoregulatory models. Furthermore, this work aims at providing a general paradigm of how personalized thermal conditioning systems can be designed and implemented with regard to software design patterns, data and control flow, software and hardware choice, programming languages and libraries as well as physical structure, experiment design, testing and validation. The system is assessed with regard to validity of measurement methods, prediction accuracy, user comfort and overall power consumption. This comprises the following experimental studies: contactless skin temperature measurement using thermal infrared is examined by comparing the method with conventional sensors and through variation of tracked measurement points. Heating and cooling of localized body regions using on-off control, proportional-integral-derivative (PID) control and model predictive control (MPC) is tested using a thermal mannikin. In addition, a user study is conducted which examines all proposed concepts, namely training of individual models through user feedback and contactless skin temperature measurements as well as subsequent control of local actuators under changing room air temperature conditions.