Study And Simulation Of Anaerobic MBR Treatmenting Low-strength Recalcitrant Printing And Dyeing Wastewater

Through physicochemical treatment of printing and dyeing wastewater belongs tolow-strength recalcitrant wastewater, which using conventional biological treatment processis difficult to reach the discharge standard. According to the low-strength recalcitrantwastewater of water quality and water features, this paper puts forward the anaerobic MBR（AnMBR） this technical route processing physicochemical treatment of low concentrationdifficult degradation of printing and dyeing wastewater solutions. The experimental resultsshow that AnMBR treatment of low-strength recalcitrant wastewater effluent can meet thenational wastewater discharge standard. This paper mainly on AnMBR treat low-strengthrecalcitrant wastewater of printing and dyeing wastewater treatment effect and support vectormachine （SVM） in the sewage treatment process modeling application.Based on the AnMBR reactor treatment of low-strength recalcitrant wastewater, theexperimental results show that: AnMBR technology in the treatment of the physical andchemical treatment of low-strength recalcitrant wastewater is feasible in technique, and theability to long-term and stable operation. Influent wasewaterCOD mean519.1mg/L, effluentCOD have significantly reduced, the biggest removal rate is86.8%, the minimum removalrate was72.6%, and the average removal rate at83.2%. The effluent wave is small, in69-119mg/L between changes, an average of86.7mg/L, basic to achieve discharge standard.Influent wastwarer BOD₅in71.1mg/L²38.2mg/L between changes, the average125.1mg/L. Effluent BOD₅average of24.1mg/L, achieve discharge standard. The influentwastewater B/C is between0.15and0.43changes, an average of0.24. The effluent water wasbetween0.16and0.46changes, an average of0.28. After AnMBR process B/C increases alittle. Influent wastewater SO₄^2-in254mg/L⁵95.9mg/L between change, an average of461.2mg/L, effluent water SO₄^2-range is60.4mg/L⁷6.2mg/L, an average of56.4mg/L.Water SS changes fluctuation is bigger, in74mg/L³34mg/L between, an average of201.6mg/L. Compared with water, AnMBR of supernatant fluid SS changes are much more stable,in53mg/L²54mg/L between change, an average of148.6mg/L. Through the membraneroom no obvious SS were detected.SVM on the basis of the grid search method, genetic algorithm and particle swarmoptimization （pso） algorithm three methods used to model parameters optimization. Set up the water COD, BOD₅and sulfate simulation prediction model. For COD simulation gridsearch method to find the best training model （R=0.895535, mse=0.00399773）, the modelin the prediction data of overfitting phenomenon, lead to GA and PSO algorithm to optimizethe model do not use, this also is machine learning the inevitable defects, is due to highdimension, function is simply stated high disaster. For the simulation of the water BOD₅found grid search method to training （mse=0.00378534, R=0.89389） has the best accuracyand prediction model of PSO has the best accuracy （mse=0.0245828, R=0.299681）. For thesimulation of the water sulfate that PSO method for training （mse=0.001, R=0.998487） hasthe best accuracy, and forecast model GA has the best accuracy （mse=0.0181346, R=0.496873）. Through the three model the comparison of three methods that grid search methodfor small sample data model predictions have certain advantages, but also should avoid thegrid search method brings a fitting phenomenon, can not blindly pursue high precision.For COD model using GA and PSO optimization in a overfitting phenomenon, thispaper proposes the use of rough set attribute reduction methods to solve because dimensiondisaster appear a model fitting phenomenon. In order to determine the model promotioneffect, different HRT of two hundred set of data model, to obtain the very good simulationresults SVM in the sewage treatment process has great development potential.