Complex glass facade modelling for Model Predictive Control of thermal loads: impact of the solar load identification on the state-space model accuracy

Above and beyond improving the efficiency of the building envelope and the energy supply
system, the demand-side flexibility in terms of load shifting and peak reduction are vital factors in further
increasing the share of volatile renewable energy sources. The thermal activation of building components,
like floors and ceilings, enables the cost-effective potential for short-term energy storage to fulfil these
requirements. In order to exploit the storage capabilities of active building systems, a reliable model predicted
control (MPC) approach is required. However, primarily if a large glass façade element is utilised,
the appropriate modelling of solar loads is critical for an effective MPC operation. Hence, based on a
dynamic building simulation tool, a characteristic map for the solar load prediction of a glass façade
system in combination of external venetian blinds was generated to enhance the state-space model approach
for the MPC algorithm. The comparison with a conventional state-space model approach shows
the integration of a detailed characteristic map can only marginally improve the prediction accuracy.
The additional information required from the glass façade manufacturer and the associated simulation
effort is not of substantial value. In contrast, the conventional grey box model enables an entirely datadriven
parameter identification, without the manufacturers’ data. Furthermore, the MPC optimisation
procedure, searching for the best control strategy, can be more efficient (solver-based optimisation), with
shorter computing turnaround times.