Water Quality Prediction Based On Morphological Parameters Of Internal And External Sludge

Wastewater treatment involves physical,chemical and biological processes,among the conventional water quality monitoring indicators,chemical oxygen demand(COD_cr),biochemical oxygen demand（BOD）,total nitrogen（TN）and total phosphorus（TP）are the key indicators to measure the effect of wastewater treatment.The traditional off-line analysis methods are time-consuming,and the results lag behind.Meanwhile,the method of on-line automatic monitoring is expensive and needs more time to maintenance.With the development of image analysis technology,the composition characteristics of microorganisms in sludge can be analyzed in qualitative method and quantitative method.Combined with multiple regression model,the effluent quality can be predicted in advance.In this study,the inner and outer returned sludge are taken as the research objects,and the floc morphology parameters of inner and outer returned sludge were obtained by microscope image combined with software analysis.Then,the results combine with the operation parameters of the wastewater treatment plant,the partial least squares（PLS）and principal component analysis（PCA）,radial basis function（RBF）neural network models are used to establish the effluent quality prediction models of the aerobic tank from the linear and nonlinear point of view.It can be concluded:（1）The correlation between floc morphology parameters of inner and outer returned sludge and operation parameters of sewage treatment plant are analyzed.It can be found that strong correlation between influent parameters and inner and outer sludge floc morphology parameters,but the correlation between aerobic tank temperature T and p H and inner and outer returned sludge floc morphology parameters is relatively general.In addition,a strong correlation exists in the morphological different floc parameters.（2）The PLS water quality prediction model（PLS1）and PCA-RBF neural network water quality prediction model of inner return sludge were established.The RMSE of COD_e、BOD_e、TN_e and TP_e in PLS model validation set are 1.878,0.979,1.429 and0.457,respectively.In addition,principal component analysis（PCA）is used to extract principal components,and nine principal components are used as input to establish effluent based on RBF.The RMSE of COD_e、BOD_e、TN_eand TP_e in the validation set were 3.064,1.019,1.284 and 0.621,respectively.The experimental results show that the two models are effective for the prediction of effluent quality,and can predict the content of effluent quality with high accuracy.（3）The PLS water quality prediction model（PLS1）and PCA-RBF neural network water quality prediction model of outer return sludge were established.The RMSE of COD_e、BOD_e、TN_eand TP_e in PLS2 model validation set are 1.993,1.095,1.409 and0.436,respectively.In outer returned sludge,the RMSE of COD_e、BOD_e、TN_eand TP_eare 2.833,1.081,1.031 and 0.297,respectively.Compared with the four water quality prediction models,PLS1 model has the best prediction effect for COD_e、BOD_e,while PCA-RBF in outer returned sludge neural network model is better for TN_e and TP_e.It shows that the degradation process of COD and BOD tends to be linear,while the degradation process of TN and TP is more complex and more suitable for nonlinear model.