Shortest-prediction-horizon model-based predictive control for individual offices

When employing model-based predictive control (MPC) for zone level heating and cooling systems, in many cases weather forecasts were imported to predict a thermal zone's temperature response over a time horizon. However, an office's thermal response is strongly influenced by an occupant's presence and behaviours. As illustrated through the analysis of an EnergyPlus simulation, propagation of the uncertainty introduced by an occupant's presence and behaviours into the temperature response of a thermal zone can result in suboptimal control decisions when the prediction time horizon extends beyond one hour. Results indicate that modest, yet robust to occupant behaviour, energy savings can be achieved by limiting the prediction time horizon to one hour in zone level MPC implementations. Choice of this prediction time horizon also eliminated the need for importing weather forecasts. In an effort to discuss the implementation challenges, this MPC algorithm has been implemented in a commercial controller to automate a ceiling radiant panel heater and a variable-air-volume (VAV) terminal unit serving to a west-facing office in Ottawa.