Nonlinear Model Predictive Control for Energy Efficient Cooling in Shopping Center HVAC

Details

16:10 - 16:30 | Tue 20 Aug | Lau, 6-209 | TuC2.3

Session: Building Control

Abstract

In this paper we present a novel approach to control a shopping center HVAC system which significantly reduces the amount of energy spent on cooling. The HVAC system considered is for a section of a Danish shopping center, including central ventilation, fan coil units and a chiller delivering cooling. The system is modeled using a grey-box RC-equivalent approach and identified parameters using measurement data extracted directly from the Building Management System from several days of live operation. From a comparison with measurements it has been concluded that the model is usable for the purpose of control design. An optimal control problem to minimize total cooling effort by manipulating central ventilation supply temperature and chiller forward temperature has been posed. The intention being to shift cooling from the chiller to the ventilation unit when cooling is available through a low ambient temperature -- avoiding both heating and cooling the same air. This optimal control problem has been used as the basis for a Model Predictive Controller. For prediction purposes, input signals from the previous days have been used, exploiting the fairly periodic behaviour of the system. Simulation studies show that during heating seasons the Model Predictive Controller is capable of shifting the entire cooling load to the ventilation unit and still maintain the same performance as the nominal controller. This amounts to energy savings of 21%.