Modelling and intelligent climate control of buildings
Doctoral thesis, 2005
The main purpose of the work presented in this thesis is to examine the possibilities of different control techniques together with intelligent building technology to improve the indoor climate and/or the energy efficiency of buildings. In particular, the possibilities of measuring more variables and using them as input to the controllers are examined.
The first part of the thesis deals with different ways to obtain dynamic models for climate systems in buildings. A large part of the thesis then deals with the feed-forward control strategy and how it affects the indoor climate and the energy use. The main conclusion, drawn from the simulations in this part is that a more extensive use of feed-forward from internal disturbances could be very advantageous in many temperature-control applications. It gives better controller performance, and, at the same time, it will often reduce the energy use.
Different controllers (P, PI, PID and ON/OFF with and without dead-band) used for indoor climate control are also investigated in this thesis. The results show that simple controllers like an ON/OFF controller with dead-band or a P-controller often performs better than more advanced controllers in many temperature-control applications. Similar types of controllers have also been investigated when they are used in a demand-control ventilation (DCV) system. The results show that there is a large potential of reducing the outdoor air flow rate by using a DCV-system instead of a base/forced ventilation system. However, the differences between different controllers in a DCV-system are of less importance in these systems.
A large part of this thesis is about the problem how to develop mathematical models for prediction of the indoor temperature using linear models as well as non-linear artificial neural network (ANN) models. The results show that neural network models give more accurate predictions of indoor temperature than linear models. ANN-models have also been used for estimation of the operative temperature in buildings. It is shown that the operative temperature can be estimated fairly well by using variables which are more easily measured and that ANN-models give better estimations than linear models. Finally, neural networks have also been used in a new method for (self) tuning of PI and PID controllers. By measuring a number of points at the step-response of a process and using them as input to a successfully trained neural network, the network can estimate the PI and PID parameters with good accuracy for the same process according to well-known tuning rules for PI and PID controllers.
indoor climate control
demand controlled ventilation