Research On Prediction And Analysis Methods Of Dissolved Oxygen And Dissolved Inorganic Nitrogen In Coastal Waters

Studying the changing laws of water quality factors in coastal waters is of great significance to preventing marine ecological disasters and maintaining the ecological environment of the seas.At present,the existing water quality prediction research is mainly for non-marine environment,but lacks the prediction research for the coastal waters environment,especially the large spatial scale and long-term span.This paper takes Dissolved Oxygen(DO)and Dissolved Inorganic Nitrogen(DIN)two key indicators to characterize water quality as the research objects,adopts the water quality monitoring dataset of the coastal waters of Weihai,and analyzes the natural factors such as time and space characteristics,biochemical characteristics,physical characteristics,meteorological characteristics,and hydrological characteristics of sea water quality comprehensively,as well as human factors such as sea region functional zoning,land-source emissions,etc..This paper is based on integrated learning methods to study Seasonal prediction method of DO and DIN concentration in coastal waters.The main works of this paper are as follows.First,data preprocessing and target factor feature analysis.According to the characteristics of the water quality monitoring data in the coastal waters,we preprocess the original data with missing value processing,feature construction,feature standardization and feature selection,and analyze the temporal and spatial distribution characteristics of DO and DIN concentrations.Second,we propose a single-factor prediction method for DO and DIN concentrations in coastal water.We integrate hierarchical clustering and density clustering to get the HD clustering model.According to the temporal and spatial distribution characteristics of the target factor,we cluster the data of monitoring station,and divide the marine research region into several partitions according to the clustering results,so that the target factors in a same partition have similar changing laws.Then in each partition,we predict the seasonal DO and DIN concentrations based on XGBoost.Third,due to the close correlation between the water quality factors of the coastal waters,on the basis of the single-factor prediction method,we propose a multi-factor prediction method for the water quality factors of the coastal waters based on the correlation chain regression.This method makes full use of the correlation between the multiple output factors.Through the regression chain,the generalization ability of the prediction model is improved,which effectively improves the prediction accuracy of DO and DIN concentrations.This paper uses the monitoring data of 160 monitoring stations in the coastal waters of Weihai from 2015 to 2019 as experimental data.The experiments verify that the single-factor prediction method of coastal water quality based on HD-XGBoost is effective for seasonal forecasting of target factors with temporal and spatial distribution characteristics.Experiments also confirm the improvement effect of the multi-factor forecasting method of coastal water quality based on correlation regressor chains on the forecasting performance of DO and DIN concentration.