Algorithm Research Of Selecting Characteristic Wavelength In COD Based On UV Spectroscopy

According to reports from the China Environmental Protection Agency: China's major rivers are heavily polluted,and water pollution is becoming increasingly prominent.In order to control the water pollution of the water resources,it is imperative to monitor the water quality.Only by monitoring and analyzing the water quality can we determine the pollution degree of the water quality and choose the optimal treatment plan.The main projects of water quality monitoring in China include chemical oxygen demand(COD),BOD,total organic carbon(TOD)and Turbidity.Among them,chemical oxygen demand(COD)is an important indicator reflecting the content of reducing pollutants in organic matter in water,and it is also one of the comprehensive indicators for evaluating the degree of water pollution.There are many ways to measure COD,which can be roughly divided into chemical and physical methods.Among them,the chemical method mainly measures the COD content by reacting various chemicals with organic reducing pollutants in water,and the precision is high,but the time is long,the secondary pollution is easy,and the rapid on-line detection is difficult.The physical mode mainly measures the absorption characteristics of certain wavelengths by using infrared visible light spectrum and ultraviolet visible spectrum.It is characterized by no chemical reagents,no maintenance,no secondary pollution,high precision,and can be monitoring online in real time.The use of ultraviolet spectroscopy to detect water quality COD is an important development direction in the field of COD detection in the academic world.However,the UV spectral water quality COD detection sensor still has the problem that the measurement accuracy needs to be improved.Therefore,this paper will improve the algorithm of the UV spectral water quality COD detection sensor based on this.In the algorithm research of COD detection by UV-visible spectroscopy,in the problem of establishing a regression model using single-wavelength ormulti-wavelength,the early spectroscopy method is more common to establish a regression model using the absorbance at 254 nm wavelength.Since the regression model is established by using a single wavelength,the stability of the obtained data is poor,so most scholars choose multi-wavelength to establish a regression model in the later stage.On the multi-wavelength selection,there are two problems:1.Simply select the multi-column wavelength with the largest correlation coefficient or fuse the extracted feature wavelength with the data pre-processing;2.How to select characteristic wavelengths efficiently and conveniently.In order to solve the above problems,this paper will use the combination of genetic algorithm and partial least squares method,particle swarm algorithm and partial least squares algorithm,and improve the hybrid genetic particle swarm optimization algorithm and partial least squares method to combine the three algorithms to select the characteristic wavelength.And the partial least squares regression model established by selecting the characteristic wavelengths of these three algorithms is compared with the partial least squares regression model of the traditional feature wavelength selection.The actual data shows that the model is established by selecting the characteristic wavelengths by three algorithms,which not only can achieve the mean square error,but also the goodness of fit exceeds 99%,the average error is 0.476853,and the average error rate is 2.3.%,has very good properties.