A control-oriented hybrid model for a direct expansion air conditioning system.

In addition to indoor air temperature, indoor air humidity is also an important parameter for building up a thermally appropriate artificial indoor environment. However, introducing air humidity as a controlled parameter in addition to air temperature would significantly increase the difficulty to develop a control-oriented model for building air conditioning systems. This is particular true for direct expansion (DX) air conditioning (A/C) systems, whose operational parameters are highly coupled and behave non-linearly and influenced by the controlled parameters, i.e., air temperature and humidity. Neither physical modelling approach nor artificial neural network (ANN) modelling approach could solely satisfy the requirement, in terms of accuracy and sensitivity, for simultaneous control of air temperature and humidity using a DX A/C system, without any inadequacies. In this paper, a hybrid modelling approach is proposed, which uses the physical modelling approach to simulate the performance of evaporator for accurately catching the cooling and dehumidification processes under various working conditions and uses ANN to simulate all other components of a DX A/C system for reduced calculation efforts. By such a hybrid modelling approach, the advantage of simplicity of an ANN-based sub-model could be utilized and the disadvantage of it that do not allow to accurately extrapolate beyond the range of the data used for training/estimating the model parameters could be avoided.