Control methodologies could lower energy demand and consumption of heating, ventilating and air conditioning
(HVAC) systems and, simultaneously, achieve better comfort conditions. However, the application
of classical controllers is unsatisfactory as HVAC systems are non-linear and the control variables
such as temperature and relative humidity (RH) inside the thermal zone are coupled. The objective of this
study is to develop and simulate a wavelet-based artificial neural network (WNN) for self tuning of a
proportional-derivative (PD) controller for a decoupled bi-linear HVAC system with variable air volume
and variable water flow responsible for controlling temperature and RH of a thermal zone, where thermal
comfort and energy consumption of the system are evaluated. To achieve the objective, a WNN is used in
series with an infinite impulse response (IIR) filter for faster and more accurate identification of system
dynamics, as needed for on-line use and off-line batch mode training. The WNN-IIR algorithm is used for
self-tuning of two PD controllers for temperature and RH. The simulation results show that the WNN-IIR
controller performance is superior, as compared with classical PD controller. The enhancement in efficiency
of the HVAC system is accomplished due to substantially lower consumption of energy during
the transient operation, when the gain coefficients of PD controllers are tuned in an adaptive manner,
as the steady state setpoints for temperature and RH are better reached in a shorter period of time.
The multi-zone analyses are suggested for future work.