Publication Date


Date of Final Oral Examination (Defense)


Type of Culminating Activity


Degree Title

Master of Science in Mechanical Engineering


Mechanical and Biomechanical Engineering

Major Advisor

John F. Gardner, Ph.D.


Thad B. Welch, Ph.D.


Hoda Mehrpouyan, Ph.D.


Aykut Satici, Ph.D.


Increased electricity production from intermittent renewables presents a challenge to utilities. Since the grid has little ability to store energy, fluctuations in solar and wind generation require either an increase in generation from expensive sources or a reduction of demand. Demand Response (DR) programs focus on the latter and are designed to increase grid flexibility by allowing grid operators to modify when or how customers use electricity. For residential customers, this typically means shedding load during periods of high demand through a central controller temporarily shutting off air conditioning (AC) compressors. This type of DR can cause spikes in demand after the units come back online.

As the communication and computational capabilities of smart meters and smart thermostats grow, so does the potential to create more decentralized approaches to DR programs. This thesis presents novel thermostat on/off criteria that rely on limited peer to peer communication between a network of residential thermostats. Agent based modeling (ABM) software was used to simulate the emergent behavior that results from thermostat interactions. To demonstrate the benefit of communicating thermostats, the criteria were tested as a means to improve the response following an AC shut off DR event and as an alternative to such events.

The introduced criteria, by sharing only the state of neighboring compressors, improved the overall demand profile following a DR event by reducing peak demand up to 21%. However, it was also found to increase the number of cycles an AC unit experiences by 36%, which can reduce its lifetime. Additionally, the stability implications of this approach are explored.