Research On Data-driven Intelligent Operation Method For Urban Sewage Treatment Processes Based On ASM Model

With the rapid development of the economy and the improvement of residents’living standards,the problem of water environmental pollution in China has gradually deteriorated.In order to build a green and sustainable development-oriented social environment,relevant departments put forward higher requirements for sewage plant emission indicators.However,at present,in order to meet the discharge standard,the sewage plant often adopts extensive and conservative management and lacks intelligent management scheme.In addition,the sewage treatment process is a multivariable,nonlinear,and strong coupling system,which is highly complex.Therefore,the sewage treatment management scheme faces numerous challenges.In recent years,the mainstream sewage treatment system management schemes can be divided into two categories:the ASMs model and the data-driven model.The former is modeled by many reaction processes of sewage treatment,which has good expression effect,but the scheme highly depends on a large number of model parameters,and the measurement method is complex,so it is difficult to deal with the complex sewage treatment process in time.The data-driven model excavates the sample characteristics and laws hidden in the data from the perspective of statistical theory to improve the intelligent management level of sewage treatment.However,this scheme makes highly abstract modeling of sewage treatment process and ignores the impact of biochemical mechanisms on sewage treatment.Therefore,based on the pilot system,this paper proposes an intelligent management scheme combining ASM 2D model and data-driven model to improve the level of sewage treatment management.This paper focuses on three aspects:First,the ASM model construction and data enhancement method based on A~2O process.Firstly,the ASM 2D mechanism model is used to model the sewage treatment process via fully investigating the operation status and basic parameter design of the pilot system.The operation data in June 2021 is obtained from the cloud platform of the pilot system.After preprocessing,it is input into the ASM 2D model for simulation calculation,and the sensitivity of each parameter is experimentally analyzed to verify the model.Secondly,the variation range of inlet water quality of the sewage plant in recent years is investigated.Based on the engineering operation experience and the research results of other scholars on the control conditions with the same process,the Influent index concentration and control conditions are divided into a series of gradient ranges,and each index is combined in the form of full arrangement.Among them,the water inflow and the temperature of each unit are set as constant values.The experimental results show that Y＿H、Y＿PAO、μ＿AUT、K＿NH has a great influence on the effluent index.After adjusting the parameters through the experiment,the simulation accuracy of effluent COD,NH₄⁺-N,TP,and TN are 93.94%,87.79%,86.68%,and91.43%,respectively.More than 2000 supplementary data were obtained in the data enhancement experiment.Second,data-driven sewage treatment process modeling.Considering the complexity of the sewage treatment process and the difficulty of many parameters of the ASM 2D model,to overcome this challenge,a data-driven model is introduced to characterize the mapping relationship between influent and effluent of the pilot system to abstract the sewage treatment process.Firstly,the collected data are preprocessed,the influent index concentration and control conditions are transmitted to the model input module as model input values,and the training set and test set are divided.Secondly,the bootstrap sampling module is used to repeatedly sample the training set data to generate multiple regression trees.Then read the test set data,make multiple predicted values according to the data characteristics learned in the training process of each tree,and take the average value of the predicted values of all regression trees as the output value of the random forest regression model,so as to realize the prediction of sewage effluent index.This study uses five common data-driven models:random forest regression model,MLP model,SVM model,KNN model,and Ada Boost model.The experimental results show that the random forest regression model has the best performance in this research scenario.In addition,the performance of the trained random forest regression model on the expanded data set is better than that on the engineering data set.Third,the ASM model and random forest regression model collaborative driving method.The supplementary data set is combined with the engineering data set to form the enhanced data set,which is used as the input of the random forest regression model,and the engineering data is used as the test set of the model,to verify the guiding role of the enhanced data set on the actual engineering data of the pilot system.The test results show that the average accuracy of the model for the prediction of effluent COD,NH₄⁺-N,TP,and TN are 97.066%,87.908%,92.442%,92.036%,respectively,indicating that the enhanced data have a certain guiding effect on the model.However,with the passage of time,the guidance effect of the enhanced data on the pilot system is weakened,and the predicted accurate removal rate of various indicators decreases in varying degrees:the effluent COD decreases from 97.066%to 83.252%,the effluent NH₄⁺-N decreases from 87.908%to 66.837%,the effluent TP decreases from 92.442%to 83.322%,and the effluent TN decreases from 92.036%to 91.792%.In this study,by increasing the weight of the new data fed back by the project in the training set,the model is retrained to improve the adaptability of the model to the pilot system.Compared with the previous model,the average prediction accuracy of effluent COD,NH₄⁺-N,TP,and TN increased by 11.022%,25.315%,13.103%,6.480%,respectively,and the prediction accuracy recovered to more than 92%,which can continue to effectively guide the operation of the pilot system.Through the investigation and analysis of the pilot system,this paper establishes a data-driven scheme based on the ASM 2D mechanism model to explore the intelligent operation scheme of the urban sewage treatment process on the premise of sewage treatment meeting the discharge standard.This study has the ability to respond flexibly to the changes of effluent quality in the future and opens up a new idea for realizing the intelligent management of sewage systems.