Study On Risk Early Warning Of Sudden Pollution Accidents In River Type Water Source Areas

In recent years,a number of major sudden water pollution accidents have occurred worldwide,not only causing huge economic losses,but also causing social instability and serious damage to the ecological environment.Due to the uncertainty of these accidents,current research only attempts to effectively prevent or actively mitigate them.In this context,this thesis mainly conducts research on anomaly detection,pollution source classification,and risk assessment and early warning for sudden pollution events in river type water sources.In the aspect of water quality anomaly detection,a data-driven model integrating improved genetic algorithm(IGA)and backpropagation neural network(BPNN)has been established.Firstly,high-precision prediction of the original water quality time series is performed using this model.Secondly,based on the residual analysis between the predicted value and the actual monitoring value,isolated points of the predicted residual vector group of water quality indicators are identified to achieve anomaly detection of water quality.Finally,a case study was conducted on the time series of three indicators(turbidity,conductivity,and dissolved oxygen)from the historical water quality monitoring data of the Longgang Phoenix Bridge Monitoring Station on the Manghe River in Yancheng City.The abnormal detection performance of the model was analyzed using the receiver operating characteristic curve(ROC).In terms of pollution source classification,a river water quality pollution source classification method based on similarity measurement of water quality indicators has been proposed.Firstly,the necessity of pollution source classification was elaborated,and methods for measuring the similarity of water quality characteristics of pollution sources were compared.Cosine similarity was selected to measure the similarity of pollution source characteristics.Secondly,a pollution source classification model was constructed,and a pollution source feature extraction method based on stacked autoencoder(SAE)was proposed.A pollution source feature library was established through hierarchical kmeans clustering method.Finally,an example analysis was conducted on the performance of the proposed feature extraction method and the constructed pollution source sample feature library based on the detection results of water quality anomalies.In terms of risk warning for sudden pollution accidents,a risk warning model based on fuzzy comprehensive evaluation and uncertainty theory has been established.Firstly,the Fuzzy Analytic Hierarchy Process(FAHP)with constraints was used to calculate and allocate the weights of various indicators in the system.Secondly,the risk scores of each indicator were quantified using Monte Carlo water quality simulation and expert scoring.Through the fuzzy comprehensive evaluation method,the risk assessment and warning level of sudden pollution accidents in rivers were determined.Finally,a case study was conducted on the risk warning model established against the background of the 2005 Harbin water pollution incident.In summary,this article establishes a data-driven model based water quality anomaly detection method for real-time identification of water quality anomaly events,and judges the types of pollution sources based on the characteristics of anomaly detection results.Finally,a comprehensive risk warning indicator system is established to assist relevant departments in emergency response and decision-making.