Development of an adaptive neuro-fuzzy method for supply air pressure control in HVAC system

Abstract

An adaptive neuro-fuzzy (ANF) method is developed for the supply air pressure control loop of a heating, ventilation and air-conditioning (HVAC) system. Although a well-tuned PID controller performs well around normal working points, its tolerance to process parameter variations is severely affected due to the nature of PID controllers. The ANF controller developed overcomes this weakness. The controller design involves 1) the constructing a Takagi and Sugeno-type fuzzy rule-based system, 2) employing the BP learning algorithm combined with the least squares method to optimize the membership function (MF) parameters, and 3) adding a secondary loop to ensure control performance. Compared with PID and original fuzzy logic controllers, simulation results show that the ANF controller performances are comparable to the well-tuned PID controller under normal conditions. It, however, exhibits a much improved robustness when the system encounters large parameter variations. It is also expected that the ANF method developed can be easily extended to other control loops in HVAC systems.