Deep Purification Of The Urban Sewage Plant And Its Simulation Study Based On BP Neural Network

In the background of the global water shortage, water recycling has become oneof the important ways to solve the current water crisis. The effluent of the townsewage treatment plant as the second water resource of cities，which is not only thewater stability，but also very good economic benefits. With the work thoroughdevelopment of China’s the12th five-year plan "energy conservation and emissionsreduction”, the public is of common concern on the stable operation of sewagetreatment plant and water quality standards. In October2005, the state environmentalprotection administration of the provinces, autonomous regions and municipalitiesdirectly under the environmental protection agency issued a about the implementationof the urban sewage treatment plant pollutant discharge standard "notification file（environment and development[2005]no.110）:In order to prevent water eutrophication,urban sewage treatment plant effluent discharged into the national and provincial keyriver valleys and lakes, reservoirs, determining the closed, semi-closed waters, itshould carry out the level of A standard; in other regions, it can be performed level oftwo standards, and according to the local actual situation, gradually increasing of therequirements. of sewage discharge.Based on the investigation to the north of a town sewage treatment plant,analysis of two effluent water quality of the plant, using biological aerated filterprocess for advanced treatment of tail water. Because the water quality of lowconcentration, so use of activated sludge and biofilm method; Follow-up by MathWork’s MATLAB7.0as the numerical platform, using BP neural network toolbox,establish the mapping model of inlet parameters and indicators of water, in order toprovide theoretical guidance and production experience for the online monitoring.Follow-up by Math Work’s MATLAB7.0as the numerical platform, using BP neuralnetwork toolbox, establish the mapping model of inlet parameters and indicators ofwater, in order to provide theoretical guidance and production experience for theonline monitoring.The experiment consists of two parts. The first part: study on biologicalaerated filter to process two effluent effect of low concentration. By setting different hydraulic retention time-1H,1.8h,2.5h,3.5H, different ratio of gas andwater-1.5:1,2.5:1,3:1,3.5:1, different temperature5¹4℃、15²0℃、21²9℃,comprehensive comparison of the pollutant removal efficiency，to determine theoptimum process parameters of the system:Hydraulic retention time2.5h, the ratio of gas and water3:1, temperature above20℃. Under the optimal process parameters，the average removal rate of CODcr、NH₃-N and SS were35.2%、57.2%、57.9%;average effluent concentrations were43.0mg/L,4.1mg/L,5.6NTU.The effluent CODcr, NH₃-N, SS value can meet the "integrated wastewater discharge standard"（GB18918-2002） in a class A standards,to achieve the purpose ofd eep treatment of wastewater. The two part: simulation of secondary effluent based on BP neural network. Adjust aeration biological filter at the optimum technological parameters, selection of continuous monitoring operation data, X=[T,CODcr, SS, NH₃-N, MLSS, DO] as the input vector, Y=[CODcr, NH₃-N,SS]as the output vector. Application of BP artificial neural network toolbox,the effluent model of BAF process was established，at the same time, the simulationvalue and the actual value was verified. When input layer neurons is6, the hidden layer neurons11, output layer neurons3,learning rate0.01,training number5000,the maximum relative error for predicting effluent CODcr concentrationwas0.95%,the MARE was0.58%;the maximum relative error for predicting effluent NH₃-N concentration was5.8%,the MARE was2.8%; the maximum relative error for predicting effluent SS concentration was7.7%,the MARE was5.3%.The average relative error below for predicted value and actual valueof three indexes were under5%,so the method could be achieved better prediction accuracy.Finally,by weight analysis, it was explored the weight contribution of each factor. For prediction model of effluent CODcr, followed by:DO>T>CODcr>MLSS>NH₃-N>SS;for prediction model of effluent NH₃-N, followed by:T>NH₃-N>DO>CODcr>SS>MLSS; for prediction model of effluent SS,followedby:SS>MLSS>DO>NH₃-N>CODcr>T.DO and T occupied the main position of weightcontribution,which is consistent with the experimental results of first part.Preliminary test technology provided the production practical experience fordeep treatment of urban sewage plant and improving effluent water quality in thefuture; the latter simulation study was based on optimum conditions of the BAF, itprovided a method to predict the effluent quality for guiding normal operation ofsewage treatment plant, and provided a decision-making basis for realizing of tailwater discharging standards.