Research On Classification Of Characteristic Contaminant In Water Distribution System Using Conventional Water Quality Parameters

With the increasingly severe challenges of urban water supply security,it is of great importance to building up an online monitoring system for the municipal water supply pipeline system.After detecting the presence of pollutants,to provide warning information about pollutant characteristics and the emergency solution,it is necessary to identify the types of pollutants.There are many kinds of substances which may cause water pollution,and some do not have targeted detection instruments.In the face of this situation,this paper studies the relationship between pollutants and conventional water quality parameters.Based on this,this article studied the classification of contaminants.The main contents and innovative points are summarized as follows:(1)After the study of relevant response laws between conventional water quality parameters and some heavy metals,organic salts and inorganic salts pollutants and analyzing the amplitude variation characteristics of different monitoring indexes,this paper put forward the technical framework for the classification and identification of different pollutants by using the different characteristics and their combined information of different pollutants.(2)This paper studied the method of determining the types of pollutants by measuring the similarity between the conventional water quality parameters.First,prediction deviations are obtained within AR and then clustered within K-Means method,finally using the similarity measure method for the identification of the contaminant.Focusing on the problem that the response amplitude of conventional water quality parameters is affected by the concentration of pollutants,the paper analyzed the cosine distance which mainly measures the angle between water quality parameter vector so that it is less affected by the amplitude and possesses the better performance in the pollutants classification and identification.By comparing cosine distance with the Euclidean distance and Mahalanobis distance,it is verified that cosine distance has the best classification results as the similarity measurement method.(3)Given the nonlinear and inconsistent changing trend between conventional water quality parameters and pollutant concentration and the lack of training samples,this paper proposed the extension multi-classification model based on SVM for pollutant classification.Taking into account the misclassification rate of contaminants is high in the initial injection phase,the probability of classification is introduced,and the difference of the samples was distinguished by studying the maximum classification probability and the standard deviation of the classification probability.Finally,some detailed performance comparison such as advantages and applicability between the similarity measure method and SVM multi-classification model has been analyzed.(4)Basic functional modules of the pollutant classification software have been developed based on the classification method of pollutants and the mixed programming technology of C#and MATLAB.The function of the software includes the classification of pollutants,the dynamic update of the feature library,the management of classification algorithms,and the display of classification results.At present,it has been applied in the simulation of water quality monitoring system based on the laboratory.