Abstract
In low energy buildings, the effect of internal and solar gains on heat balance of rooms is large. As a result, the heating systems, designed assuming steady-state conditions with no heat gains, are over-dimensioned for most of the heating period. This poses a challenge for room-based control systems, especially for thermostatic valves, but also for PI controllers. Using over-dimensioned room units might result in room temperature fluctuations. For finding solutions to this problem by using simulations, correct modelling of the control system together with the room is crucial. Therefore, the aim of this research was to determine the challenges that occur while matching measured and simulated temperature profiles and test the effect of PI control parameters on the calibrated model control accuracy. The experiments were carried out for the underfloor heating system of a test building. The building was simulated in IDA-ICE software and calibration via minimising root mean square error of energy consumption in GenOpt was carried out. The PI parameters were fit by optimisation with objective to simulate the measured temperatures accurately. The effect of the optimized PI parameters was determined by comparison to IDA-ICE default parameters and parameters from Cohen-Coon method.
Highlights
It is common practise to over-dimension the heating systems in buildings for the sake of safety margins
On the two remaining parameters, the gain and integration time, we ran the optimisation by minimising the sum of temperature root mean square errors (RMSEs) via GenOpt, the results are shown in Table 5 together with the default parameters as well as parameters calculated from Cohen-Coon method
Obtaining the PI control parameters through optimisation proves successful as the RMSE results show a small improvement even compared to the parameters calculated from the Cohen-Coon method
Summary
It is common practise to over-dimension the heating systems in buildings for the sake of safety margins. The current load calculation standard does not take into account the internal gains making the estimated power demand even higher [1]. Recently literature discusses self-learning, adaptive and predictive controllers, still many of the “smart” controllers make use of the classical controllers to reach and maintain the calculated setpoint they suggest [3]. Tuning the simple controllers has been a relevant topic for decades [3] and current self-learning PI controllers make use of it by running several tests and calculations to auto-tune the parameters online. In building energy simulations, varying the parameters is not common as ideal systems are used or parameters are set to default values and the effect is often under-estimated [4]. Finding better parameters than default ones, requires expert knowledge in control theory and calculations for identifying suitable parameters
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have