The performance of three adaptive controllers has been compared with experi
mental studies on a VVS-400 Heating/Ventilation System. The control algorit
hms are Generalized Predictive Control (GPC), an adaptive PID Controller (A
PC) based on Dahlin algorithm and the VVS-400 Heating/Ventilation System lo
cal controller-a Fuji PID auto-tuner. For the on-line identification module
of the first two adaptive control algorithms, the recursive least-squares
estimation technique is implemented. It is demonstrated that the two model-
based approaches outperform the PID auto-tuner. Moreover, the adaptive PID
controller compares favorably with a much more sophisticated GPC. (C) 1999
Elsevier Science Ltd. All rights reserved.