Simple Tool to Evaluate the Performance of Model Predictive Control of Space Heating in the Early Stages of Building Design

Simulation-based studies have indicated the significant theoretical potential of using Model Predictive Control (MPC) in space heating systems to leverage building thermal mass for demand response through energy flexibility. Recent reports on applying MPC in hydronic heating systems in real residential buildings suggest this potential can be realized in practice. As evidence of MPC’s energy flexibility potential grows and technologies become more robust, building designers will increasingly need to assess MPC’s impact on building performance during the early design stages using thermal building performance simulation tools.

A typical MPC scheme for space heating involves an optimization algorithm that determines the optimal heating control action based on minimizing an objective function with appropriate constraints. Simulation-based studies of MPC concepts often use co-simulation, where buildings or zones are modeled in a physics-based environment like EnergyPlus, and the heating system is controlled by an MPC defined in another digital toolbox such as MATLAB, using data exchange software like the Building Controls Virtual Test Bed. The control model in the MPC can be a Resistance-Capacitance (RC) model calibrated with data from experiments on the virtual building or a black-box model generated using system identification techniques.

While co-simulation methods can assess MPC performance in the building design process, they require advanced skills and are time-consuming, making them impractical for the rapid changes and explorative nature of early design stages. Consequently, there is an emerging need to define a simulation setup that allows building designers to quickly and reliably assess MPC as a space heating control strategy early in the design process. Currently, there are no reports in the literature addressing this need.

To address this gap, this paper presents a concept enabling building designers to evaluate MPC for space heating control during the early design stages. A discrete-time linear time-invariant MPC scheme, where the control model is described using block matrices, is proposed. This system can be optimized for all hours of a year in one go, allowing for fast exploration of how different building design variables affect MPC performance. The scheme was implemented in an existing combined thermal and daylight simulation tool. Results from a case study, featuring design variables such as thermal capacity, window-to-wall ratio, and glazing type, illustrate how the proposed setup enables time-efficient assessment of MPC performance in building design proposals during the early stages of building design.