Modeling of dissolved oxygen concentration in activated sludge process based on generalized dynamic fuzzy neural networks

This paper is concerned with modeling of dissolved oxygen concentration in the activated sludge wastewater treatment process, using the generalized dynamic fuzzy neural network modeling (GDFNN) method, to predict the change of dissolved oxygen concentration. This method uses an elliptical basis function (EBF) as its fuzzy membership function, as to the width of it will be adjusted according to the importance of input variables. At the same time its features such as learn online, self-organization and trim of rules can improve the accuracy and generalization ability of the dissolved oxygen concentration model. Finally, the effectiveness of the GDFNN is illustrated by comparing with dynamic fuzzy neural network (DFNN) and BP neural network. Simulation results show that the GDFNN modeling method can improve the accuracy of the dissolved oxygen concentration model effectively, and has good generalization ability.