ANN IDENTIFICATION APPLYING TO MODEL STRUCTURE OF THE DIRECT-FIRED LIBR ABSORPTION CHILLER

摘要

One drawback of direct-fired LiBr absorption chiller lies in its excessive fuel consumption. Good control strategy is an effective way to improve the performance of the chiller systems and the associated equipment. There are two basic but quite different principles for model construction: the "a priori" approach and the "a posterior" approach. The former one is also known as "physical modeling", and the other one "identification" by which the models are derived from experiment data. This paper proposed using artificial neural networks to identify the nonlinear MIMO system such as absorption chiller system. In order to have informative data and to be able to find a suitable model structure, this paper presented using the physical model to eliminate the redundant or unnecessary variables in inputs and outputs.