Prediction And Analysis Method Of Nitrate Content And Chemical Oxygen Demand In Coastal Sea Areas

At present,eutrophication of water body is the most prominent problem that puzzles the ecological environment of coastal waters in China.Eutrophication is due to the excessive content of nitrogen,phosphorus and other nutrients in the water,which makes the input and output of nutrients unbalanced,and furthermore destroys the stability and function of the water ecosystem.Eutrophication affects the sustainable development of marine ecological environment and marine economy seriously.Therefore,how to effectively predict and prevent eutrophication in coastal waters has become a hot issue for scholars at home and abroad.According to the evaluation method of water eutrophication,nitrate and chemical oxygen demand are the key indicators affecting water eutrophication.The variation lavws of these two factors indirectly reflect the occurrence of eutrophication.Therefore,water eutrophication can be predicted by the variation of nitrate and chemical oxygen demand(COD).Based on the water quality monitoring data obtained by Weihai Marine and Fisheries Monitoring and Disaster Reduction Center,the progressive gradient boosted regression trees(GBRT)and time series analysis method are utilized in this thesis to model and analyze the nitrate and chemical oxygen demand content in the coastal waters.and obtains the variation laws of nitrate content and chemical oxgen demand.The main work of this thesis incluaes three aspects.First,Water quality,data preprocessing.The original monitoring data are preprocessed by abnormal value processing and missing value interpolation.Among them,abnormal value processing is based on the 3? rule,and the water quality monitoring data satisfying the conditions are screened out within the threshold range of normal distribution.The LIN interpolation algorithm is used to interpolate the missing values,and the missing values are calculated according to the correlation between the monitoring values of the adjacent days.Second,to divide the functional areas into nitrate and chemical oxygen demand.This method divides the whole water area according to the characteristics of geographical location and sea area function,and predicts and analyses the small watershed separately after dividing,which effectively improves the analysis and prediction effect.Third,time series analysis method was used to predict COD.Through empirical mode decomposition of the periodic signals of COD,multiple sub-signals are obtained,and autoregressive integrated moving average model(ARIMA)is used to model and process the sub-signals respectively.The values of integrated prediction are obtained as the final results.The water quality monitoring data of Weihai coastal waters from 2005 to 2018 is collected as the experimental data in this thesis.The experimental results verify the feasibility and validity of the machine learning model proposed for regularity analysis in coastal waters.