This paper deals with buildings indoor air temperature control problems. It specially aims applications concerning actual heating devices testing and control strategies analysis. In both cases, accurate but low-dimension dynamical models representing the thermal behaviour of the building are required. Knowledge models satisfy accuracy requirements but they usually include a great amount of differential equations to solve. Using model size reduction techniques becomes thus essential for control applications. Different model size reduction methods are presented and tested in PID control and optimal control problems. Their performances are shown through an example of application.