Sensible Home: Micro-environmental control through wearable personal sensorsEPA Grant Number: SU836940
Title: Sensible Home: Micro-environmental control through wearable personal sensors
Investigators: Wang, Julian , Fan, Howard
Current Investigators: Wang, Julian , Fan, Howard , Feng, Yanxiao , Yakkali, Sai Santosh , Duan, Qiuhua
Institution: University of Cincinnati
EPA Project Officer: Callan, Richard
Project Period: February 1, 2017 through January 31, 2019 (Extended to January 31, 2022)
Project Amount: $75,000
RFA: P3 Awards: A National Student Design Competition for Sustainability Focusing on People, Prosperity and the Planet - Phase 2 (2016) Recipients Lists
Research Category: P3 Challenge Area - Chemical Safety , Sustainable and Healthy Communities , P3 Awards
The purpose of this project is to develop a novel micro-environmental control (MEC) system consisting of a wearable module, a central control module, and an outdoor weather module. The personal wearable module (PWM) can 1) sense the occupant’s physical (e.g., movement, posture) and physiological (e.g., stress, mood) behaviors, and 2) monitor their micro-environmental conditions. These signals are sent to the central control module (CCM), which determines the occupant’s indoor location and behaviors, and accordingly adjusts the building systems to autonomously optimize for the occupant’s micro-environmental conditions; this is accomplished through a multi-objective genetic algorithm embedded with comfort and an energy prediction models. Users can adjust the control algorithm when necessary, typically by adjusting setpoints. The outdoor weather module (OWM) supplies the external wind and air quality information used for potential natural ventilation control. This objective of this research is to achieve a working engineering prototype which can bring the human-in-the-loop of building system controls, and eventually lead to new advances in building energy efficiency and indoor comfort.
The long-term goal of this research is to implement the proposed MEC system into residential buildings with a healthcare purpose and goal of saving energy. Therefore, two tasks completed in service of this Phase II fund application were the identification of a target area and a general quantification of the opportunity. We determined that the ideal entry market would be dwellings designated for a senior population, because 1) the U.S. DOL ATUS data showed that people aged 65 and over tended to spend more time in their rooms than did other age populations; on average, this value was approximately 15.1 hours per day.30 2) Senior populations were found to have significant demands regarding indoor comfort and health.31 The overall objective of Phase II, then, is to further develop our proposed micro-environmental control system for this specific population and move it from laboratory to practice. The experimental implementation of this system is planned for a senior living residence, in collaboration with a local non-profit continuing care retirement community called Maple Knoll, an industry partner in smart building systems named Kens, Inc., and University of Cincinnati UCRI which is an independent not-for-profit center focusing on the technology transfer to market.
Specifically, the technical objectives of this Phase II project include the following. First is to focus the integrative comfort model, based on the Phase I resultant model, on the senior population (age 65 and above) through national and international standards, as well as handbooks and research reports on the senior population’s indoor comfort. Second is to develop a reliable behavioral recognition model. Use of our customized wearable sensors for data collection and the data analysis approach employed for Phase I have proven effective. Therefore, the same methods will be followed in Phase II, but with larger samples (40 subjects) and a longer duration for data collection (seven days). Third is to establish an energy model based on the energy simulation method explored in Phase I. The energy model will be used in the central computing process for control optimization. Four is to finalize the design of the PWM, based on the initial design and sensor architecture generated in Phase I. The industrial design researchers on our student team will focus on this task, and also conduct a user-friendliness study. Five is to develop an effective and efficient optimization control logic for the proposed MEC system. The behavioral recognition, comfort, and energy models will all be integrated and processed through this multi-objective control logic. We seek to develop a technique that uses simple hardware, without compromising the handling of nonlinear and complex models. Last is to move our designed system from the laboratory and into a real living environment, the Maple Knoll senior community, and connect it to the building system and senior residents. Through this activity, we will be able to accurately access the energy savings and comfort level improvements through actual energy use data, subjective responses, and a computational simulation.
Built upon the theoretical demonstration in Phase I, the final goal of Phase II is to improve and verify our concept and prepare it for implementation, in an effort to benefit individual indoor comfort (people), promote economic growth in smart buildings and healthcare (prosperity), and build energy efficiency (the planet).
In Phase I of this project, four objectives or tasks were achieved as expected.
First, various indoor comfort models were investigated by reviewing national and international standards, handbooks, and literature related to human factors. An integrated comfort framework was formed, and an Environmental Condition Generator tool was programmed using Matlab. This generator tool randomly outputs proper environmental conditions (temperature, humidity, illuminance, air speed, etc.) by inputting different human factors (behaviors). These resulting environmental conditions will be further integrated into the energy models and central computing of the CCM.
Second, the occupant comfort desire was determined to depend upon the user’s indoor activities and environmental conditions. We utilized an online database – the American Time Use Survey (ATUS) – accessible from the U.S. Department of Labor (DOL) website to categorize indoor environment related human behaviors into eight types, with distinct thermal and visual comfort and air quality requirements. Consequently, a pilot human behavioral recognition study was conducted in order to determine ways of recognizing these eight categories of behavior. Six subjects, including two faculty investigators and four student researchers participating in this project, were selected. A customized Microsoft band (wearable sensor) was utilized for the data collection. The six subjects were asked to stay at home and wear the band for three days; they documented their behavior, hour by hour, in a diary form, referencing the above-listed eight categories of behavior. By using a clustered data mining method, we initially set up the behavioral recognition model using wearable sensor data (skin temperature, heart rate, and motion). However, we will improve and reshape this process in Phase II of the project, if funded.
Third, building upon the previous two tasks, we explored a new energy simulation method for estimating energy savings that uses the MEC system in residential buildings. A U.S. DOE residential prototypical model – a single family detached house – with Cincinnati, Ohio weather file, was adopted as the Reference Model. The model integrated with the control logic of our proposed MEC system was named the MEC Model. The Reference Model complied with the latest energy standard: the 2012 edition of the International Energy Conservation Code (IECC). On the one hand, with regards to the outcomes of this energy estimation in relation to the reference model, our proposed system was shown to improve the building’s HVAC and lighting energy use by ~23.1% and ~11.3%, respectively. At the overall building energy use level, the energy savings were approximately 22.3% (78.4 MBtu, or 49 KBtu/sq ft). On the other hand, regarding the indoor comfort improvement, this energy simulation showed a 41.1% thermal comfort index increase and 29% visual comfort level increase.
Four, multiple sensors recording data related to heart rate, skin temperature, air temperature, relative humidity, and ambient light were combined as a single “wearable module” and installed on a pole in a space; a weather station was also mounted on the exterior of the space. A laptop was used as a central station. After gathering data on various conditions, we tested the sensor IC selections, overall function and operation, hardware connections, data communication, etc. The outcomes of this task enabled us to understand the potential sensor dimensions, functions, diagrams, and user interface. Building upon this knowledge, several design options for the personal wearable module (PWM) were proposed from the early design perspective.
The use of human factors and corresponding comfort requirements enabled the building control to continuously adapt its systems in response to the real needs of the occupants; the preset schedule, with steady temperature and lighting setpoints, that is used in conventional building environmental controls provides only low comfort levels and high rates of energy use. This Phase I project proved effective at reducing energy consumption, with an estimated annual reduction in energy of ~22.3%. Using this proposed MEC concept also increased individual thermal and visual comfort levels by approximately 41.1% and 29%, respectively. However, there were three limitations found, which require further study. 1) Unspecified population: None of the preliminary Phase I studies involved specific types of occupants; instead, they used small samples. However, many studies have shown that different populations (delineated by age, gender, and other factors) may have different indoor comfort needs. Therefore, our behavioral models and patterns are, at this point, insufficient. 2) Insufficient behavioral recognition model: In the Phase I project, we focused on proving that wearable sensor data (skin temperature, heart rate, motion) can be used to recognize human behaviors; in this, we were successful. However, the initial behavioral recognition model was based on a small sample size (six subjects) and a short duration (two days). 3) Lack of control logic: The whole system was proposed to achieve optimal building system settings for both individual comfort and energy savings, which is a characteristic multi-objective optimization problem. The control logic design with an optimization algorithm was an important component of the whole system, but the Phase I work didn’t include this element. 4) Testing in real living environments: After Phase I, the proposed MEC system was still at the laboratory stage and had not been tested in actual living environments. Real living environments will likely involve more building information such as appliances, construction types and conditions, and external conditions, as well as more human factors including actual living patterns and interactions. Testing in real living environments would verify and also further improve our proposed MEC system.