A recent webinar instalment for Sim4Blocks was given by Donal Finn and Dimitrios Kapetanakis from University College Dublin (UCD) on Building Energy Prediction using Reduced Order Modelling.

Ranked 161st in the Times Higher Education World University Ranking, UCD’s major strategic research themes include energy and environment, argi-food and health.

The agenda for the third webinar of the Sim4Blocks project included general methodology, a Sim4Blocks case study undertaken by partners HFT, the simplification and evaluation of Resistance-Capacitance (RC) models and future project plans. The RC approach is a methodology aimed at evaluating energy consumption in buildings.

The webinar outlined how the RC approach could be used with regards to the demand response (DR) objectives of the Sim4Blocks project. In the current project they are used because it is easier to simplify a representation of a buildings heat transfer. The balance between building energy gains and losses can be converted into a series of RC equations.  Further integration of the various energy conversion systems is also possible. These equations are quicker to solve and are capable of being integrated within larger power systems analysis scenarios. However, one disadvantage is that their accuracy requires in-building calibration.

The general methodology that UCD are using to understand how the RC model can create a simplified building model is made up of four key steps:

  • Develop and calibrate RC models based on building data
  • Evaluate the performance of RC models for various prediction horizons
  • Further calibration of the RC models will be necessary as further data becomes available
  • Integrate with Model Predictive Control models being developed elsewhere in the project

To gather the data from their case study – a detached three storey building supplied by partners HFT in Würtenrot – UCD used the computer method Particle Swarm Optimisation (PSO) algorithms, whereby after a number of repeat processes a group of variables have their values adjusted and with every adjustment the value gets closer to the desired target. The case study involved a single heat pump and underfloor heating while HFT provided weather, indoor temperatures and ventilation data. From mid-October to mid-February the case study looked at the heating model in the building set with a specified RC configuration and error predictions were determined.

Future work will look at different order RC circuits and calibration using measured data.