A neural network based temperature control for air handling unit

Abstract

In this paper, a neural network (NN) based temperature control scheme is presented for cooling coil of Air Handling Unit (AHU) to overcome the unknown system dynamics and its inherent nonlinearities. First of all, by utilizing the standard thermodynamic formulas, the cooling coil is modeled by a discrete-time uncertain nonlinear system instead of a linear one, in order to accommodate its different behaviors under various cooling loads. Thereafter, a two-layer NN is utilized to approximate the unknown dynamics within the system. The online tuning algorithm of the NN weights is also provided with no offline training needed. Consequently, the identification process of the cooling coil unit is obviated. Theoretical results show the stable tracking performance of the close-loop system, while the effectiveness of this method is demonstrated under simulation environment. © 2011 Springer-Verlag.