Study On A Multiplex Spectral In-situ Water Quality Monitoring Technology

Water resources security is the guarantee of national ecological security and people’s health.With the development of the social economy,countries around the world have become more and more strict in the supervision of water resources security,with related policies for water resources protection being continuously introduced.However,water pollution caused by incidents such as furtive discharge,secret discharge,and malicious discharge has occurred from time to time,causing unnecessary pollution in many river basins in China.Most of the water quality detection technologies such as manual sampling laboratory analysis,shore station online,and mobile monitoring(environmental vehicle,ship)use chemical methods for detection and analysis,which are inefficient,but requiring a large number of chemical reagents that are prone to a secondary pollution.A high-efficiency in-situ water quality monitoring technology based on spectral technology can realize real-time online multi-information monitoring.Spectral water quality monitoring technology,as a purely physical monitoring method,has high detection efficiency,without any chemical reagents and secondary pollution,making it a green and pollution-free rapid water quality detection technology.This paper focuses on the research of spectral feature extraction and pollution source identification of water samples,in which,Spectral information processing is one of the core technologies.The water sample spectrum as a complex matrix can be simplified into a superimposed combination of multiple single pure substances.This paper first studies the spectral information of a variety of single pure compounds,and proposes applicable discrimination methods to describe the spectral characteristics of compounds,including spectral barycenter,dynamic piecewise integration,and effective absorption bandwidth,all of which are used as the core parameters of water quality monitoring.Next,the(ARMA)prediction model was constructed using the time series of spectral features,and the anomaly monitoring model abandoned the traditional fixed threshold.With the use of confidence interval of predicted residual as the abnormal monitoring threshold in the ARMA prediction model employs,experiments show that the detection rate of the prediction model with the spectral barycenter as the monitoring value reaches 98%,and this threshold is not limited by the monitoring water source.Subsequently,a method for tracking and determining suspected pollutants based on DTW distance is proposed,which solves that the difficulty in comparing different dimensional spectra.Finally,in the in-situ monitoring experiment of the lake,the water quality anomaly monitoring algorithm based on the spectral barycenter can accurately detect the abnormal water quality changes,and trace the pollutants that caused the anomalies.In summary,this paper further improves the method of water quality spectral information processing by studying the spectral characteristics of environmental water samples,time series prediction models,and suspected pollutant determination method.The experimental results indicate the fast calculation speed and accurate determination of pollution in water quality monitoring and abnormal tracking,pollution identification,et.al.in this paper,which greatly helps to improve and expand the detection capability of in-situ online water quality monitoring technology of spectrometry.