The Structure Self-Organizing Algorithm Of Fuzzy Neural Networks And Its Applications In Wastewater Treatment Process

Activated sludge wastewater treatment method is among the main methods in the treatment of industrial organic wastewater and secondary municipal sewage at present. It is a kind of effective means, which consumes the pollutant in wastewater by microorganisms'metabolism. Due to the violent change of influent and the complexity of microorganisms'growth, the activated sludge process has characteristics of multi-variable, strong coupling, high non-linearity, big-lag and indetermination, which results in laggard modeling and control of wastewater treatment process. For the sake of enhancing the quality of output water, decreasing the costs of wastewater treatment and ensuring the highly efficient operation of wastewater treatment system, the research of intelligent modeling and control methods of wastewater treatment process can not only be valuable in theory and reality, but also provides useful reference for the modeling and control of other complex systems with nonlinear and big-lag features.Fuzzy neural networks (FNN) are hybrid systems that combine the theories of fuzzy logic and neural networks, thus can make effective use of easy interpretability of fuzzy logic as well as superior learning ability and adaptive capability of neural networks. It is widely used in areas of adaptive control, adaptive signal processing, nonlinear system identification, pattern recognition, and so on. The design of fuzzy neural networks consists of structure identification and parameter identification. However, conventional neurofuzzy design techniques address only the parameter identification, the structure of network is established according to the knowledge of experience. On account of the evident drawbacks showed in the approach, a new self-organizing fuzzy neural networks (SOFNN), where a flexible structure partitioning can be made according to different objects, is proposed in the paper.On the basis of a thorough analysis of the present achievements, the author studied the structure self-organizing algorithm of fuzzy neural networks and its applications to modeling and control of activated sludge wastewater treatment system. The main accomplishments of this paper are as follows:(1) A new structure self-organizing algorithm of fuzzy neural networks is proposed, while comprehensive comparisons with other approaches are made, through examples of nonlinear function approximation and nonlinear system identification, to demonstrate that the new algorithm is superior in terms of compact structure and learning efficiency.(2) Based on a deep analysis of the inherent mechanism of activated sludge wastewater treatment system, a forecast model of output-water quality of the system is established with the proposed algorithm. Simulation results show that the model is effective with high performance, compared with fuzzy neural network model.(3) A self-organizing fuzzy neural network controller, based on the proposed approach, is applied to the control of dissolved oxygen. Simulation studies show the controller is of good performance and effective.The research work in the paper may be valuable, to some extent, for expanding the intelligent modeling and control methods of wastewater treatment in our country.