Practical implementation and evaluation of model predictive control for an office building in Brussels

A model predictive control (MPC) has been implemented in a medium-sized office building in Brussels, Belgium. This paper presents the implementation of the controller and the measured performance in comparison with the default, rule-based control (RBC). The building has two floors and a total size of 960 m(2). The controllable system is the hybrid heat production consisting of two air/water heat pumps and a condensing gas boiler. The practical situation does not allow controlling end-units in the different zones of the building. The MPC makes use of a Modelica grey-box control model resulting from a system identification with monitoring data. The paper covers the monitoring, model identification, forecasting of disturbances, state estimation, formulation and solving of the optimal control problem (OCP) and transmission of the control signals. The performance is evaluated on a daily basis based on analysis of heating degree days, thermal comfort, energy costs and primary energy consumption. The results show that the model predictive controller is able to provide a similar or better thermal comfort than the reference control while reducing the energy costs by more than 30%. This is due among others, to a better use of the heat pumps and an adapted hot water supply temperature. (C) 2015 Elsevier B.V. All rights reserved.