A robust form of self-tuning controller that has been designed to replace conventional PID controllers in a number of HVAC applications is described° The controller is based on a version of the Generalized Predictive Control algorithm that uses default values for most of its parameters and requires selection of only one commissioning parametec the control sampling interval Special jacketing software has been developed to implement a set of expert rules, which are not application dependent, but which supervise the operation of the on-fine parameter estimator and the calculation of control action. This ensures satisfactory operation of the controller in most situations.

The performance of the controller is evaluated under demanding conditions using a detailed computer simulation of the central air handler in a building. A componentbased computer simulation package (HVACSIM + ) was used to examine the behavior of the controller in both the zone and supply air temperature control loops. Results demonstrate the robust behavior of the self-tuning controller in a variety of circumstances and suggest that its performance is superior to that of a manually tuned PI controller.