Development of artificial neural networks for performance prediction of a heat pump assisted humidification-dehumidification desalination system

In this study, the application of data-driven methods for performance prediction of a heat pump assisted humidification-dehumidification (HDH-HP) desalination system was investigated for the first time. Although HDH-HP desalination systems have been widely studied both theoretically and experimentally, the application of data-driven models as a powerful predictive tool has not yet been investigated in these systems. To fill this gap, three data-driven models (MLPANN, RBFANN, and ANFIS) were applied using 180 experimental samples. The gain output ratio (GOR), heat transfer rates of the evaporator Q̇e, and evaporative condenser Q̇c, were considered as outputs. The results indicate that the MLPANN model is superior in predicting all target parameters showing R2 values of 0.915, 0.995, and 0.988 for GOR, Q̇e, and Q̇c, respectively. Further, the ANFIS model performance was shown to be weak for predicting GOR. Finally, a comparison was made between the experimental heat transfer rates, MLPANN model, and compressor polynomials. The predicted values using the MLPANN model were found to be in excellent agreement with experimental data, possessing a MAPE of 0.48% and 0.77% as compared to predicted values by compressor polynomials with MAPE of 9.53%, and 3.3%, for Q̇e, and Q̇c, respectively.