Research On The Short-term Eutrophication Prediction Based On Big Data Of Observation Network Multi-sensor

With the rapid development of economy and society and the aggravation of human production activities,the pollution of water resources in China becomes increasingly serious.Eutrophication of water body is one of the most serious problems of water pollution,which poses a great threat to the stability and diversity of water ecosystem.Water eutrophication prediction can help us know the quality and pollution status of water environment in time,and take necessary actions for comprehensive treatment of water body according to its possible development trend in the future.Based on the online monitoring system of tuanjie estuary water quality,this study carried out real-time online monitoring of the water quality in the coastal waters,analyzed and predicted the water pollution status,especially the eutrophication problem through monitoring data.First of all,this study focuses on the overall structure of the water quality monitoring system and various water quality parameter monitoring methods,and puts forward some improvements to the design of the system.A water quality monitoring system architecture based on multi-sensor observation network is proposed,which can realize more stable collection and transmission of water quality data.A scheme of storage and data analysis based on big data technology is proposed for mass water quality monitoring data,which provides some reference for mass water quality data processing.Secondly,this study for water quality data preprocessing method in detail,different from other types of data,the index coupling relationship between water quality data,which makes the processing water quality data to deal with individual characteristics exist certain problems,on the basis of the research of the data set missing values and outliers common problems put forward the corresponding solution.In order to reduce the redundancy of data features,factor analysis method was used to reduce the dimension of water quality data set,and eight water quality parameters,such as water temperature,PH,dissolved oxygen,chlorophyll a,turbidity,chemical oxygen demand(COD),phosphate and nitrate,were determined as factors influencing eutrophication.Finally,in view of the large data scenarios of water quality data processing and forecasting,this study selected depth(within DNN)and the integrated neural network learning algorithm XGBoost return projections for chlorophyll a,according to the concentration of chlorophyll a,to reflect and predict eutrophication condition,forecast interval choice for 1 day for the efficiency of the water environmental governance provides a certain support.Based on their performance on the test data set,both models have higher prediction accuracy,but the maximum relative error of the XGBoost model is 9.7%,nearly half of the maximum relative error of the DNN model of 18.6%.