COMPARISON OF MULTI OBJECTIVE GMDH-TYPE NEURAL NETWORK AND BAYESIAN BELIEF NETWORK IN THE PREDICTION OF TREATED WATER TURBIDITY. CASE STUDY: GREAT WATER TREATMENT PLANT IN GUILAN PROVINCE

Enough water with proper quality is necessary for life. Drinking Water Treatment Plants (WTPs) have to provide high quality drinking water in the shortest possible time with minimal costs. In this paper, Factors affecting the process for removal of water turbidity using the Response Surface Methodology (RSM) were firstly identified and then GMDH-type Neural Networks and Bayesian Belief Network (BBN) have been used for modeling and prediction of treated water turbidity; using input-output data set. To validate the proposed model, a case study was carried out based on the data consisted of 700 sets obtained from Guilan‌WTP. For modeling, the experimental data obtained from the operation unit were divided into train and test sections (70% for training and 30% for testing). The predicted values were compared with those of experimental values. The determination coefficient of the predicted values for the two BBN algorithms consist of EM and GD, and GMDH model were 0.9388, 0.9196 and 0.97095, respectively. The GMDH model performed better than the BBN model in predicting treated water turbidity dosage.