Soft-sensing Research For Water Ammonia Nitrogen Based On A Self-organizing Recurrent Rbf Neural Network

With the acceleration of social urbanization and industrialization,the demand for industrial production processes and urban water using is increasing,and the subsequent problems of water pollution have also increased.This has brought a great deal of influence to the survival and development of human beings and the ecological balance of society.The ammonia nitrogen is one of the important pollutants in water resources.It can be used as an important indicator to evaluate the quality of water bodies.The ammonia nitrogen exceeding the standard will cause eutrophication of the water and cause environmental pollution.Therefore,in order to reduce the harm caused by exceeding the ammonia nitrogen concentration,the measurement and research of ammonia nitrogen in the sewage becomes very important.Due to the fact that the measurement method of ammonia nitrogen in the actual sewage treatment plant has the disadvantages of cumbersome operation,low measurement accuracy,long lag time and high instrument maintenance cost,the ammonia nitrogen content is difficult to achieve online detection.Therefore,for the real-time detection of effluent ammonia nitrogen in wastewater treatment process,an effluent ammonia nitrogen soft-sensing model based on recursive orthogonal least squares algorithm for self-organizing recursive RBF neural network(ROLS-RRBF)is proposed.Besides,the development of an effluent ammonia nitrogen soft-sensing intelligent system was realized,and the accurate prediction of effluent ammonia nitrogen was realized.The research work of this article is mainly divided into the following points:(1)The research on self-organizing recursive RBF neural network designing based on recursive orthogonal least squares algorithm.This paper presents a structural design method of self-organizing recurrent RBF neural network based on recursive orthogonal least squares algorithm.Firstly,according to the independent contributions of hidden layer nodes to the output layer neurons,singular value decomposition is used to determine the optimal number of hidden layer neurons,and the dynamic adjustment of the structure of the recursive RBF neural network is implemented.At the same time,an adaptive optimum steepest descent learning algorithm is used to train the parameters of the recursive RBF neural network to ensure the accuracy of the network.Finally,the experimental simulation is depicted in this paper.Compared with other self-organizing methods,the simulation results show that the self-organizing recursive RBF neural network based on recursive orthogonal least squares algorithm can achieve higher prediction accuracy with a more streamlined structure,which lays a solid theoretical foundation for the establishment of the effluent ammonia nitrogen soft-sensing model in this paper.(2)The research on effluent ammonia nitrogen soft-sensing model based onROLS-RRBF neural network.In order to predict the the effluent ammonia nitrogen content in wastewater treatment process in real time,firstly,the mechanism of the ammonia nitrogen participating in the reaction was analyzed,and the principal component analysis was used to select the auxiliary variable as the input,and the effluent ammonia nitrogen as the output.Then,an effluent ammonia nitrogen soft-sensing model based on ROLS-RRBF neural network was established.Finally,in order to verify the validity of the soft-sensing model,its model was applied to the actual sewage treatment plant.The experimental results show that the soft-sensing model can effectively realize the on-line prediction of effluent ammonia nitrogen.(3)The development of a soft-sensing intelligent system for effluent ammonia nitrogen.This paper designs and develops a effluent ammonia nitrogen soft-sensing intelligent system,which mainly includes: user management module;homepage module of ammonia nitrogen soft-sensing system;related information module of ammonia nitrogen soft-sensing;neural network model module;second-order anaerobic-anoxic-oxic process module.In the process of designing and implementing the system,firstly,Visual Studio 2010 software is used to complete the interface design of the system.Then,the SQL Server 2008 database is used to store user information and auxiliary variables and other data,and the mixed programming techniques of C# language and Matlab is used to achieve the training and prediction of effluent ammonia nitrogen.Finally,through the information transmission among various modules such as user management module,data processing module,neural network model training and forecasting module,the predicting value of effluent ammonia nitrogen is outputted and displayed,and the purpose of visualization of the soft-sensing system interface is achieved.