model predictive control, supervisory control, peak power, retrofit, air conditioning
This article describes a low cost retrofit technology that uses collective control of multiple rooftop air conditioning units to reduce the peak power consumption of small and medium commercial buildings. This retrofit technology uses a model predictive control to select an operating schedule for the air conditioning units that maintains a temperature set point subject to a constraint on the number of units that may operate simultaneously. A proto-type of this new control system was built and deployed in a large gymnasium to coordinate four rooftop air conditioning units. Based on data collected while operating this proto-type, we estimate that the cost savings achieved by reducing peak power consumption is sufficient to repay the cost of the proto-type within a year. Moreover, it is possible to reduce the cost of this proto-type technology by a factor of at least six and thereby create a retrofit package that pays for itself within two months of operation. The effectiveness of the control in this demonstration strongly suggests that widespread deployment of the proposed technology could significantly reduce peak demand originating with small and medium buildings. Over 27% of the energy used by most small and medium sized commercial building is dedicated to air conditioning units, and most of these units continue to rely on simple, uncoordinated controls that independently maintain the temperature for their assigned sections within the building. One consequence of this uncoordinated control is that in the peak heating and cooling seasons it is almost certain that all of the air conditioning units within a building will operate at the same time. This causes large peaks in the power used for temperature control, and these peaks can be significantly reduced by the proposed retrofit technology. This reduction in peak demand benefits the utility company by reducing the need for expensive peaking plants and their associated infrastructure, and it benefits the building owner by reducing electricity costs. These advantages and the short payoff period suggest a significant commercial potential for the proposed control technology.