Design And Application Of Soft-Sensing Model For Ammonia Nitrogen Based On Radial Basis Emotional Neural Network

With the improvement of urban civilization and industrial level,the demand for water resources is increasing,which causes an increasing shortage of water resources,a large amount of sewage is discharged.Thus,the urban water pollution is more serious.In order to improve the utilization rate of wastewater resources to make the effluent quality standard,the treatment of wastewater from urban wastewater treatment plants needs to be enhanced,and the regeneration and use of wastewater resources should be promoted.Among the many pollutants causing water pollution,ammonia nitrogen is a key water quality indicator that will cause eutrophication and environmental pollution of water,and its exceeding emission can seriously damage water resources and menace the health of people,so reducing the emission of ammonia nitrogen is one of the main tasks of urban wastewater treatment plants.However,in the process of urban wastewater treatment,the actual removing nitrogen biochemical reaction is more complex,which is a nonlinear system with time variability and uncertainty.At the same time,there is noise in the sampling data,traditional chemical reaction detection methods are difficult to timely,convenient and accurate to realize online measurement.At present,the radial basis emotional neural network(RBENN)has been widely used in dealing with complex nonlinear modeling problems because it adds Gaussian function to the inputs,which can eliminate the effect on interference to a certain extent by mapping the inputs into high-dimensional space,and has the advantages of strong nonlinear approximation ability and better anti-interference performance.Therefore,in order to deal with the problem that the concentration of ammonia nitrogen in effluent water is difficult to detect in the urban wastewater treatment process,this thesis proposes a soft measurement model method for the concentration of ammonia nitrogen in effluent water based on the RBENN,which achieves an accurate prediction of the concentration of the ammonia nitrogen in effluent water.The key research works of this thesis mainly contain the points are as follows:1.SBPSO ammonia nitrogen characteristic variable selection method based on NMI and SU:To deal with the problem that the characteristics of auxiliary variables can be difficult to select due to the large variety and complex relationship of effluent water quality indicators from the municipal wastewater treatment process,a sticky binary particle swarm optimization(SBPSO)ammonia nitrogen characteristic variable selection method based on normalized mutual information(NMI)and symmetric uncertainty(SU)is designed.Firstly,a particle population initialization mechanism based on SU is designed,and the intrinsic characteristic information of the sample data is obtained,thus forming a higher-quality initial population.Secondly,a feature deletion strategy is proposed,which reduces the amount of computation in the iterative search process.Then,a particle fitness evaluation function is constructed,which guides the search direction of particles.Finally,a particle optimal location update rule is formulated,and a subset of auxiliary variable features with strong correlation and low redundancy with water ammonia nitrogen indicators are obtained.The simulation results show that the proposed feature selection method can automatically screen the auxiliary variables with strong correlation with effluent ammonia nitrogen.2.RBENN design method based on ADw-CLPSO algorithm:Aiming at the difficulty in determining the weight parameters and the network scale according to the actual task,which leads to the low generalization performance of the model,this thesis designs the algorithm called adaptive inertia weight adjustment strategy based on comprehensive learning particle swarm optimization(ADw-CLPSO).Firstly,a particle inertial weight adaptive dynamic adjustment strategy is designed to adjust the flight parameters of each particle,and improve the quality of knowledge.Secondly,a particle-variable dimensional learning mechanism(PVDLM)is established on the basis of the algorithm.Then,the designed algorithm is applied to RBENN to realize the dynamic adjustment of the model structure.Simulation experiment results show that compared with other algorithms,the RBENN is able to determine the learning parameters and improve the prediction accuracy of the RBENN model under a compact network structure.3.Soft measurement model of ammonia nitrogen based on ADw-CLPSO-RBENN:In view of the problems that the ammonia nitrogen concentration in effluent of urban wastewater treatment process is unstable,the source analysis of pollutants is complex,and it is difficult to establish an accurate model,which lead to the difficulty of achieving an online measurement of water ammonia nitrogen concentration,an ADw-CLPSO-RBENN based on soft measurement model for water ammonia nitrogen is designed and constructed.Firstly,NMI and SU based on SBPSO ammonia nitrogen feature selection algorithm,the auxiliary variable subset for input model is obtained.Then,the ADw-CLPSO algorithm is used to determine the suitable model structure of RBENN,and the parameters of the model are tuned simultaneously to build a soft measurement model.Finally,the prediction performance of the model for effluent ammonia nitrogen concentration is verified by using the actual production data of a real urban wastewater treatment plant.The simulation results show that the structure of the established soft sensor model is relatively simple,and compared with the experimental results of other soft sensor models,it can have stronger generalization performance,and can achieve accurate effluent ammonia nitrogen concentration prediction.4.APP development and design of ammonia nitrogen monitoring system for urban sewage treatment process:To realize the engineering application of the designed soft-sensor model,relying on the “Thirteenth Five-Year Plan for Water Pollution Control and Governance Major Scientific and Technological Special Subproject-Development and Research of the Big Data Platform for Water Environment Management in the Beijing-Tianjin-Hebei Region”,an APP software of ammonia nitrogen monitoring system in urban sewage treatment process based on IOS platform is designed.The main functions include the user identity management system,the visual display of historical and forecast data of water quality parameters and the query of pollution information in relevant areas.Firstly,in the process of design and coding,Swift is used to write the client interface of the software to realize the visual application of the effluent ammonia nitrogen soft-sensing system.Then,Java is used to write the server interface for the client to call,and technologies such as Matlab and My SQL database are used to obtain relevant data,which realizes the integration and management of data.The system APP is highly integrated and has a good visualization effect,which helps the staff of the relevant management departments to understand the general information of key water quality parameters in time,and check the regional pollution status so as to make corresponding treatment in time.