Efficient management of energy in buildings necessitates the accurate evaluation of the air conditioning (AC) system performance. Ideally, the system must be operated with its cooling capacity matching the cooling load demand, and with maximum possible performance. This paper presents a simplified method that can predict cooling capacity in operative AC installations with limited input information. It is developed using an artificial neural network (ANN) with Bayesian regularization. The training data are generated by numerical simulations of operating scenarios representing the real system operation. The refrigerant temperatures at the inlet and outlet of the evaporator and the condenser are selected as inputs for the proposed method to predict the cooling capacity in relevant operating scenarios, thereby eliminating the need for a flow meter and facilitating implementation on operative systems. The ANN model is developed to capture the performance of different systems by using a data normalization method. The ANN prediction is tested on both simulation scenarios and experimental data with different nominal capacities. The results show that the ANN model can successfully predict cooling capacity variations using limited input parameters with RMSE and ΔQ-e,rel 0.09 kW and 3.99%, respectively.
ASJC Scopus subject areas