Study On COD Detection Method And Modeling Of Water Quality Based On UV-Vis Spectroscopy

In recent years,with China's rapid economic growth and accelerating pace of urbanization,the problem of water pollution has become more and more serious.This problem has become one of the most serious problems that our country or even the whole world should pay attention to,for it causes a lack of water resource.As an important indicator to evaluate the degree of water pollution,the Chemical oxygen demand(COD)can characterize the concentration of organic matters in water.The Ultraviolet-visible(UV-Vis)spectroscopy is a green detecting technology with a short cycle,which can avoid secondary pollution of water and can realize on-line detection.In this paper,we have designed a detecting method based on the UV-Vis spectroscopy to detect the COD of water,and with the detecting results the prediction model of the water's COD has also been constructed.The main research can be divided into the following aspects:1.Spectral system and spectral acquisitionIn order to detect the COD value of the water,we develop a UV-visible spectrum COD detection system.The preparation and detection of the potassium hydrogen phthalate standard solution have been realised,and the UV-visible absorbance spectroscopy data has been collected.2.Research of spectral data preprocessing technologySince the original spectral data may be affected by a large amount of noise,so it is necessary to select a denoising method which is able to reduce the loss of the real data information as much as possible during the denoising process.Wavelet analysis can meet this requirement.By using the wavelet function db8,a 5-layer wavelet decomposition is carried out on the original spectrum,and then the soft threshold method is used for quantization.The reconstructed spectral curve of the water COD is very smooth,which means that the noise of spectral signal has been effectively eliminated.After the use of wavelet denoising,there still exists problems of spectral information redundancy and multiple collinearity.The principal component analysis is used to reduce the dimension of the spectral data,which can effectively eliminate redundant information,retain useful feature information and improve the efficiency of machine learning.3.Study on COD prediction model of water qualitySince there exists a complex nonlinear relationship between UV-visible spectral data and the water COD,traditional mechanism modeling methods can't be used.Water quality COD prediction model can be established by BP neural network for predicting the value of the water quality COD effectively.In order to improve the prediction accuracy,the improved whale optimization algorithm is used to optimize the BP neural network parameters,and the water quality COD prediction model based on MWOA-BP neural network is established.The prediction results show that the prediction accuracy of the model is higher and can be applied to the prediction of COD.4.Optimization algorithm improvementAs basic whale optimization algorithm have many shortcomings,such as a slow convergent speed and low accuracy of convergence,an improved whale optimization algorithm(MWOA)is proposed.The MWOA mainly studies the population initialization mechanism and the nonlinear adaptive weighting strategy.The simulation results show that the improved algorithm can maintain the initial population diversity during the optimization process.Besides,it has better convergence speed and convergence precision.