Using a multi-objective optimal design of GMDH type neural networks to evaluate the quality of treated water in a water treatment plant
Fereshteh Alitaleshi, Allahyar Daghbandan · Desalination and Water Treatment · 2019
ABSTRACT In this research, a model based on the multi-objective optimal design of GMDH I -type neural network is proposed for evaluating the quality of treated water. To validate the proposed model, a case study was carried out based on the data sets obtained from Rasht Water Treatment Plant (WTP), Guilan, Iran. For modeling, the experimental data obtained from the laboratory and operation unit were divided into training and testing groups (70% for training and 30% for testing). After modeling, the predicted values were compared with the ones obtained from the experimental values. The determination coefficient of the predicted values for the two data sets of GMDH model (laboratory and operation unit) were 0.9905 and 0.9714 respectively. Comparison between experimental and mathematical results from GMDH-type neural networks showed the success of this method.