Study On Method For On-line COD Monitoring In The Effluent Of Rural Sewage Treatment By UV-Vis Continuous Spectroscopy

With the development of social economy and the improvement of people’s living standards,human beings are paying more and more attention to the state of the water environment.In rural areas,rural sewage has not been effectively treated and there are substandard discharges,which has caused great harm to the cleanliness of the environment and human health.At present,the decentralized treatment model is generally used in rural areas,and gradually has the characteristics of small scale,scattered locations and numerous sewage treatment facilities.Therefore,it is necessary to find an economical and efficient method for real-time chemical oxygen demand(COD)monitoring of water quality in combination with the economic and environmental conditions of rural areas to ensure that rural sewage is discharged in compliance with standards.Ultraviolet-visible(UV-Vis)spectroscopy,as a fast,cheap,and non-secondary detection method,is gradually applied to the monitoring of COD,but its monitoring effect in practical applications is not ideal.Therefore,this article will take the actual effluent from rural sewage treatment facilities as the research object,aiming at improving the prediction accuracy and stability of the COD monitoring model by UV-Vis spectroscopy and realizing effective application.The main research contents of the paper include:(1)Based on the theory of UV-Vis spectroscopy,select several representative substances for experimental water distribution,and configure single-component solutions,two-component solutions,and multi-component solutions with different concentrations.The experimental study verified the good correlation between the UV-Vis spectrum absorbance of water samples and the COD concentration of the solution,and proved the feasibility of the UV-Vis spectrum method for COD prediction of water quality.(2)Taking the effluent from actual rural sewage treatment facilities as the research object,the water quality and spectral data of water samples were tested by using normal distribution test,principal component analysis,and cluster analysis.Partial least squares method(PLS),support vector machine,and back-propagation neural network were used to construct the UV-Vis spectroscopy COD prediction models.By comparing and analyzing the model prediction results,it is found that the partial least square method is most suitable for modeling.The coefficient of determination(R~2)is 0.931,and the root-mean-square error(RMSE)is11.6.(3)The wavelength selection and stepwise regression are used to perform characteristic wavelength selection because the spectral data has redundant information during the process of constructing the COD prediction model by UV-Vis spectroscopy.Under the premise of ensuring the accuracy of the model,the modeling process was simplified.The optimal wavelengths were selected at 251 nm,356 nm,and 363 nm.The R~2 of the multiple linear regression model is 0.924.(4)Use the fluorescence excitation-emission matrix regional integration(FRI)to make quantitative analysis of various components in water samples,and then use cluster analysis to identify and classify water samples.Finally,the COD prediction model is constructed for each type of water sample data,which greatly improves the prediction accuracy of the model.The R~2 of the optimal prediction model is as high as 0.994.(5)Develop COD online monitoring device based on UV-Vis spectroscopy,and complete the overall architecture design,hardware equipment selection and integration,and software function installation and commissioning.The actual water samples were used to test the performance of the online water quality monitoring device.The relative error of the prediction results is within 10%and the relative standard deviation(RSD)is less than 3.48%.