A model for the optimized control of an innovative air conditioning system with variable geometry

Abstract

In this work it was aimed to develop and optimize an artificial neural network (ANN) which accurately simulates ejector cooling cycle performance according to the operational conditions. It would also allow for the control of a spindle, located in the primary nozzle, in a way that ejector performance is maximized. First, it was aimed to optimize an ANN capable to accurately simulate the performance parameters of a refrigeration cycle and the ejector performance itself. Input variables were the operational conditions of the cycle. In this first stage, a data set with some limitations in terms of input variables domain representativeness was used including those cases that resulted in ejector failure. With this first data set it was possible to evaluate the effect of those limitations in the artificial neural network optimization. In the second stage, a new data set was presented to the selected ANN in order to assess the influence of each input parameters on the cycle performance and also on optimal spindle control. The condenser temperature was found to be the most important parameter affecting the ejector performance and so the control of the spindle position brings great advantages for optimizing the cycle performance in order of the operative conditions.