Study on the feature-recognition-based modeling approach of chillers

Abstract

With the absence of structure parameters of chillers, it is difficult to simulate the chiller performance by the conventional precise modeling approach. Thus, a novel modeling approach is proposed, which makes unknown structure parameters lumped and obtains them (i.e. the characteristic parameters, which are unique for each chiller) based on measured data of chillers. The modeling principle of feature recognition is described, based on which then the chiller models were developed. An experimental platform of chiller was established, and the relevant chiller models were established on basis of measured data. Moreover, to verify the chiller models, simulations and experiments were compared under variable water flow rate and variable water temperature conditions respectively. Results show that with the feature recognition method, the chiller models can function quickly and efficiently, and achieve a high accuracy. Compared with experimental results, relative errors of simulated performance parameters, such as refrigerating capacity and COP (coefficient of performance), are all within 10%. This study provides a new and efficient approach for chiller modeling.