Pattern Rcognition Adaptive Control.docVIP

A New Pattern Recognition Adaptive Controller John E. Seem Johnson Controls, Inc., Milwaukee, Wisconsin USA Abstract: This paper describes a novel method for automatically adjusting the gain and integral time of proportional-integral controllers based upon patterns that characterize the closed-loop response. The method was developed for self-regulating systems that can be modeled as first-order plus dead-time systems. This new pattern recognition adaptive controller (PRAC) is easy to use and provides near-optimal performance for a range of systems and noise levels. Also, the algorithm is computationally efficient and does not have large memory requirements. PRAC has successfully tuned control systems for heating, ventilating, and air-conditioning equipment in a number of buildings. Keywords: Adaptive Control, Control Applications, Energy Control, Heat Exchangers, Pattern Recognition, PI Controllers 1. Introduction The heating, ventilating, and air-conditioning (HVAC) industry commonly uses proportional-integral (PI) feedback controllers. People installing and commissioning systems do not have a long time to tune loops. Consequently, some PI algorithms use the default control parameters shipped with the controller. For some systems, the default control parameters are not appropriate and using them can lead to poor control performance. Also, many control loops require retuning during the year. Several researchers have worked on applying adaptive control methods to HVAC processes. Dexter and Haves (1989) and Jota and Dexter (1988) used the Generalized Predictive Control Algorithm (Clarke, et al., 1987) to control different HVAC systems. Dexter and colleagues concluded that the magnitude of the disturbances found in the HVAC industry causes problems when applying self-tuning and adaptive control methods. Bristol and Kraus (1984) describe a pattern recognition adaptive controller for PID controllers that is called EXACT. The EXACT control