Computer Science Colloquia
Monday, November 21, 2011
Advisor: Kamin Whitehouse
Attending Faculty: John A. Stankovic, Chair; Kamin Whitehouse, Advisor; Sang H. Son; Stephen D. Patek; Tingting Zhang, Minor representative
8:00am in Olsson 236D
Ph.D. Dissertation Presentation
A Convenient and Inexpensive Sensing System for Improving the Energy Efficiency of Heating, Cooling, and Lighting in Homes
Energy is among the most important issues in the world today and is at the heart of many concerns including the economy, the environment, and national security. Buildings are responsible for almost half of all energy consumption and greenhouse gas emission annually in the world, and are therefore essential to any energy management strategy worldwide. Most efforts to increase building efficiency focus on weatherization and equipment efficiency. For example, a November 2006 report by the President's Council for Advice on Science and Technology (PCAST) identified building insulation and energy-efficient appliances as key priorities for improved building efficiency. However, retrofitting is an expensive endeavor. Oak Ridge National Laboratory has proposed a series of deep retrofits with the potential to improve the energy efficiency by as much as 30-50%, but the total cost for each home is over $20,000 on average. The progress toward long-term energy goals is largely limited by short-term availability of capital investment. Thus, new technologies must be developed that can reduce the energy consumption of a building without requiring a large initial monetary investment.
Wireless Sensor Networks enables ubiquitous sensing and smarter control in buildings to save energy with a much smaller initial monetary cost than existing solutions that require physical modifications to the building or equipment. In this dissertation, we provide a cost-effective sensing system that saves energy in residential buildings by sensing the information about home occupancy and physical environment, predicting future conditions by analyzing patterns in historical sensor data, and automatically configuring and optimizing the building operations for energy efficiency. Our system uses motion and light sensors to reduce the two largest energy end-uses in homes: space heating & cooling and lighting, and the system automatically configures itself to obviate the cost of professional installation. It includes three major components. First, the Smart Thermostat uses occupancy statistics in a home in order to save energy through improved control of the HVAC system. This system uses a combination of long-term occupancy and sleep patterns with real-time motion sensor data to control the HVAC system. Second, SunCast is a novel sunlight prediction framework that uses historical data traces to produce a continuous distribution of predicted sunlight values. Fine-grained predictions improve the performance of natural daylight harvesting by exploiting small peaks and troughs in daylight levels. Finally, the Place-N-Play system uses a combination of motion sensors and light sensors and facilitates sensor configuration by automatically inferring the floor plan of a home and the locations of these sensors. The result is a visualization of the building floor plan that allows users to control the devices through a touchscreen interface.
This dissertation lays the foundation for next-generation buildings that will autonomously sense the building environment and strategically control building operations to achieve significant energy efficiency. The principles and approaches developed in this dissertation can be extended to commercial buildings, and can also be applied to many other aspects of building operation such as ventilation and water heating. Our technology has the potential for a large impact for its low-cost and practicality, and can be translated to market for serving an important need for building energy efficiency across the world.