Tuesday, 24 January 2017: 2:15 PM
Conference Center: Tahoma 5 (Washington State Convention Center )
Health risks associated with extreme heat events are modified by socioeconomic, behavioral, and physiological factors. In the summer of 2015, we studied the interaction between air conditioning usage patterns among the residents of two affordable housing developments in Cambridge, Massachusetts. These residents receive subsidies for their energy consumption bills. However, due to a combination of factors, such as energy savings awareness and thermal comfort, the study subjects have different patterns for using their air conditioning units. We hypothesized that late-onset use of cooling systems results in higher exposures to dangerous heat index levels. For approximately two weeks, we monitored fifty apartments where temperature, relative humidity, noise, and carbon dioxide measurements were collected. Participants living in these spaces were asked to use a wearable device measuring physical activity, sleep patterns, skin temperature, skin conductance, and heart rate. We automated the analysis of indoor temperature profiles to identify set-points and air conditioning frequency use, as well as the identification of the temperature threshold where cooling using natural ventilation transitions into mechanical cooling. We propose the combination of these separate data streams as a methodology to predict decompensation during extreme heat events by optimizing cooling regimes. Our objective is to customize thermostat control schemes based on machine learning algorithms that incorporate features from the user as well as physical characteristics of the space (e.g. orientation, floor, thermal mass, etc.). This feasible approach might reduce risks associated with heat stress and heat stroke by preserving temperatures below dangerous heat index values.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner