Predictive Modeling And Diagnosis Of Key Parameters Of Wastewater Treatment Process

With the continuous development of industrialization and urbanization,the total amount of wastewater discharged from human activities is also increasing,and the problem of water pollution needs to be solved urgently.Therefore,studying the virtuous recycling of water resources has important practical significance for alleviating the pressure on water resources.Among them,the wastewater treatment plant is the main infrastructure to solve the secondary pollution of water resources.The qualified discharge of wastewater depends on the accurate acquisition of process data and the effective implementation of process control.In the wastewater treatment plant,there are a large number of automated instruments and electrical equipment.It is easily affected by many factors,leading to frequent failures.Therefore,accurate and reliable fault diagnosis of the entire wastewater treatment process is a very meaningful research topic.This thesis takes the activated sludge process as the research object,and is based on the internationally universal wastewater treatment process benchmark simulation model BSM1(Benchmark Simulation Model no.1),based on artificial neural networks,intelligent algorithms,interval predictions and other methods,to improve the wastewater treatment process Research on key parameter prediction and fault diagnosis issues.The main research work and innovations of this thesis are as follows:(1)Aiming at the problem of difficult to accurately simulate and obtain timely and accurate data from wastewater treatment plants,this thesis uses the internationally accepted activated sludge process benchmark simulation model(BSM1)as the basis to obtain experimental data.Firstly,the mechanism characteristics of activated sludge process and the bio-reflection characteristics are studied in depth,then the bio-chemical reaction pool and the two-sinking pool of BSM1 are elaborated and analyzed in detail,and finally the accuracy and stability of BSM1 are verified by the data provided in MATLAB environment.(2)Aiming at the problems of beetle antennae search(BAS)algorithm in high-dimensional optimization problems,such as poor search accuracy and slow convergence,an improved beetle swarm optimization(IBSO)is proposed.Firstly,with the idea of particle swarm optimization,the individual search of the beetle was raised to the group search.Then,the group search strategy of Lévy flight and the step-size adaptive strategy were introduced to establish the self-adaptive search algorithm for the group of beetles,and finally verifies the reliability and validity of IBSO algorithm at high latitude through the test functions.(3)Aiming at the problem that it is difficult to realize real-time online measurement of key parameters in the wastewater treatment process,a key water quality parameter prediction model of the IBSO-Elman network is established.Firstly,the weight of Elman neural network is used as the position of foraging for cattle,and then the parameters are optimized by IBSO algorithm,so as to reduce the implicit layer node and improve the prediction accuracy.Finally,by comparing with the simulation of other predictive models,the experimental results show that IBSO-Elman has good tracking performance and strong robustness.(4)Aiming at the problem of fault diagnosis of key parameters in the wastewater treatment process,a fault diagnosis model of wastewater treatment process parameters based on multi-interval prediction is proposed.Firstly,the Bootstrap sampling method is used to train the network to obtain the total error to construct the fault diagnosis prediction interval.Then,the fault diagnosis model is constructed according to the interval prediction strategy to determine the fault type.The effectiveness and superiority of the proposed fault diagnosis method are verified by MATLAB simulation of the key parameters of the wastewater treatment process under different fault conditions.