A cost effective and non-intrusive method for performance prediction of air conditioners under fouling and leakage effect

Realistic performance predictions are required for efficient operation strategy of air conditioners. In this study, the application of a cost effective and non-intrusive black box model utilizing artificial neural networks (ANN) to predict the cooling capacity of air conditioning systems is investigated, while considering the effect of fouling and leakage that may occur after prolonged operation. The effect of various leakage and fouling combinations on the output cooling capacity were numerically simulated. The training data set is first generated for a system that is ideally operating without any fouling or leakage. The developed ANN model is tested to predict cooling capacity in “faulty” systems. The results indicated that, as long as leakage and fouling are limited below 10% and 4% respectively, the ANN model trained by the data generated with the ideal system, can predict cooling capacity with a relative averaged cooling capacity difference (ΔQ-_e,rel) of approximately 13%. Moreover, the inclusion of data with different leakage and fouling combinations in the training set enables accurate predictions of the cooling capacity of the air conditioning system during the entire timespan of its operation. It suggests that cooling capacity under the fouling and leakage phenomena can be predicted using limited input information.