Research On Accurate Treatment Method Of Wastewater COD Based On Neural Network

Dosing is an extremely critical step in the process of sewage treatment.It removes suspended matter and harmful organic matter by adding chemicals to wastewater.The amount of dosage directly affects the water treatment efficiency and effluent quality of sewage treatment plant.If the dosage is insufficient,the quality of the effluent can not meet the prescribed standards;If the drug is administered in excess,it will increase the cost of sewage treatment and bring the risk of secondary pollution.Therefore,in the sewage treatment process,precise application of medicine is crucial.However,the dosing process is very complex,involving physical and chemical processes,and is affected by many factors such as raw water quality and influent flow rate,and has the characteristics of large time delay and nonlinear change.At present,many sewage treatment plants still use manual control for drug administration,which is difficult to deal with raw water with large water quality fluctuations,because the quality and type of pollutants in raw water with large water quality fluctuations are not stable,and it is difficult to achieve accurate drug administration.Therefore,how to achieve accurate drug application is one of the most important problems for sewage treatment plants at home and abroad.In this paper,sewage dosing of a sewage treatment plant in Nanning was taken as the research background,and the research object was Fenton dosing system of the plant.More than 3000 groups of manual dosing data of the system were used as samples for analysis.In view of the characteristics of the dosing system in sewage treatment plant,such as nonlinear,large time delay,multi-variable and difficult to achieve accurate dosing,an algorithm based on artificial neural network model was proposed in this paper,which was used to accurately predict the dosing amount of Fenton reagent(ferrous sulfate and hydrogen peroxide reagents).The algorithm could effectively solve the difficulties and problems in the manual control mode.First of all,the administration data of sewage treatment plant were normalized and outlier removed.Then,the factors of raw water quality were analyzed to determine the input-output characteristics of the model.Among them,raw water flow rate,effluent pH,influent COD,effluent COD,dosage at historical time were taken as input characteristics,and two dosages of ferrous sulfate and hydrogen peroxide were taken as output characteristics.The data samples were divided into training samples,test samples and verification samples,and the multiple linear regression model,BP neural network model and RBF neural network model were established.The model was trained,tested and verified.Finally,the prediction results of the three models were compared.With higher accuracy and reliability.In order to further improve the prediction accuracy of the model,the BP neural network model was optimized by genetic algorithm.The experimental results showed that the optimized BP neural network model could reduce the prediction errors of dosage of ferrous sulfate and hydrogen peroxide by 0.0162 t/h and 0.004 t/h,respectively.The mean square error is reduced by 0.0849 t/h and 0.0041 t/h,respectively.The root mean square error is reduced by 0.1 t/h and 0.032 t/h,respectively.The prediction effect was improved.Therefore,the BP neural network model optimized by genetic algorithm proposed in this paper can accurately predict the dosage of Fenton reagent.