Date of Award

May 2015

Degree Type


Degree Name

Master of Science



First Advisor

Yingchun Yuan

Committee Members

Anoop Dhingra, Lingfeng Wang


Control, Demand Response, HVAC, MPC


One of the major challenges that building owners and operators face is maintaining a low cost of operation. In certain markets within the U.S., electrical cost varies throughout the day; it is higher during times of peak demand. This leaves the customer the incentive to cut back electrical use during peak demand periods. Since 40% of the peak electrical demand is due to the operation of the building HVAC system alone, the opportunity exists for shifting the building cooling load to off-peak hours. This can be done by pre-cooling the space, thereby using the building mass as a sort of thermal battery, which can then discharge later, alleviating the cooling load off the HVAC system during peak times. It is in this thesis that a peak load reduction strategy is presented using model predictive control (MPC). Furthermore, the system modeled in this paper is a two-zone system, each having a dedicated controller. First the problem is explored with a single, centralized MPC which calculates the optimal trajectory for the entire building. Secondly, the load reduction strategy control is distributed to each individual controller. The advantage to distributed control is the reduction of computing resources which brings with it a cost reduction on its own. Lastly, both MPC approaches are compared to the traditional PI-only control scheme. Results showed that the distributed scheme proved favorable next to the centralized MPC benchmark, and both MPC approaches produced favorable results over the traditional PI-only control.