Modeling of effluent quality parameters in a submerged membrane bioreactor with simultaneous upward and downward aeration treating municipal wastewater using hybrid models

Abstract

This research was an effort to develop hybrid multilayer perceptron and radial basis function artificial neural network–genetic algorithm (MLPANN-GA and RBFANN-GA) models to accurately predict effluent biochemical oxygen demand (BOD), chemical oxygen demand (COD), total nitrogen (TN), and total phosphorus (TP) in a submerged membrane bioreactor. The input variables of the networks were influent BOD, influent COD, influent TN or influent TP, sludge retention time (SRT), mixed liquor suspended solid, membrane permeability, and transmembrane pressure. Training procedures of all effluent quality parameters were successful for both the MLPANN-GA and RBFANN-GA models. The training and testing models showed an almost perfect match between the experimental and predicted values. Based upon the statistical analysis, results indicated that there is a very little difference between predicted and experimental values of the effluent BOD, COD, TN, and TP. The predicted and experimental values of the effluent concentrations gave a very low root mean squared error and a high coefficient of determination very close to one demonstrated high accuracy of these models to predict output variables. It became clear that the models based on the genetic algorithm (GA) were much better than those models without GA from the viewpoint of the achievement of an accurate prediction of the effluent BOD, COD, TN, and TP. The results indicated that the accuracy of all models increased when GA was applied to neural networks. The mean average error for the hybrid models varied from 3 to 8%.